GEO for Agencies Monetizing AI Search Visibility

GEO for Agencies: Monetizing AI Search Visibility ![rw-book-cover](https://wpengine.com/wp-content/uploads/2026/09/geo-for-agencies.jpg) 🔗 **Source:** [GEO for Agencies: Monetizing AI Search Visibility](https://wpengine.com/blog/geo-for-agencies/) > [!note]+ Document Notes > tags: [generative-engine-optimization, ai-search-visibility, agency-go-to-market, wordpress-agencies, vendor-marketing, content-governance] > author: Brian Gardner > source: WP Engine > reference: https://wpengine.com/blog/geo-for-agencies/ > date created: 2026-09-23 19:04 > > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > > # 🏷️ Title: GEO for agencies: Monetizing AI search visibility >  ## 🧭 Table of contents - [📝 Summary](#-summary) - [📖 Main content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning styles assessment](#-learning-styles-assessment) - [📊 Accessibility assessment](#-accessibility-assessment) - [🌍 DEIB assessment](#-deib-assessment) - [✅ Logical validity](#-logical-validity) - [🛡️ DARVO analysis](#️-darvo-analysis) - [🎯 Conclusion](#-conclusion) - [🔄 Refactor appendix](#-refactor-appendix--geo-for-agencies-monetizing-ai-search-visibility) ## 📝 Summary - **What happened.** WP Engine published a 1,684-word post by Brian Gardner on 2026-09-23: 42 paragraphs, 10 headings, 8 outbound links. Paragraphs 0 to 6 are a summary block; the body runs 1,101 words from ¶7 to ¶41. An embedded video (¶20) and two lists (the FAME pillars at ¶22, the four offerings at ¶35) did not survive capture. - **The claim.** Discovery is moving from ranked search results into conversational answers on ChatGPT, Claude, Perplexity, and Google AI Overviews. Agencies can sell Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO) as a new service line. SiteSonar, a tool built by Chandan Sharma of the agency DoodleWeb, is the recommended way to run that line: it tracks AI citations, finds competitor gaps, drafts content, and publishes to WordPress through a plugin with a human approval step. The FAME framework (Findability, Awareness, Market trust, Engagement) organizes the work. - **The mechanism.** ¶16 states it in one hedged sentence: an AI system may retrieve from the web, evaluate sources, synthesize an answer, and attach citations. ¶17 names what that rewards: entity relationships, structured information, brand authority, and content that answers the questions buyers ask. Nothing in the post cites a study, a platform document, or a measurement for either sentence. - **Why it matters.** For a reader in WordPress developer relations at a competing host, this is a complete go-to-market package in one post: a named framework, a partner-built tool, a service-packaging section, and a case study. WP Engine is telling its agency channel what to sell next quarter. The mechanism part is defensible and vendor-neutral; the product part is where every specific number lives. - **Key stakeholders.** WP Engine (publisher), Brian Gardner (byline), Chandan Sharma and DoodleWeb (tool builder, and the sole case study), SiteSonar (the product), WordPress agencies (the audience), and the four named AI platforms, none of which is quoted or linked. - **Risks and tradeoffs.** The post contains exactly one statistic: an 83% increase in "measured AI visibility" at DoodleWeb, attributed at ¶36 to "SiteSonar's internal measurement." The tool vendor measured the tool vendor's own agency with the tool vendor's own metric. No baseline, no time window, no definition of the metric. The trend claims at ¶7 and ¶8 ("a growing share," "no longer clicking") carry no source. The post does not state whether WP Engine has a commercial relationship with SiteSonar or DoodleWeb. AI-drafted "micro pages" produced continuously (¶29 to ¶30) is the content pattern Google's March 2024 scaled content abuse policy targets; the post addresses editorial approval (¶32) but not that policy. - **Bottom line.** The mechanism is worth teaching and costs nothing to adopt: structured data, clear entities, direct answers, offsite authority. The product claims and the single number are unverified in this source and should not travel without the ¶36 provenance attached. ## 📖 Main content **Two texts in one capture.** Paragraphs 0 to 6 read as a key-takeaways block: five compressed claims, a pointer to WP Engine's Technical GEO guide, and a meta-description line. Whether WP Engine or Reader wrote it, it is the only place the FAME pillars are named (¶2: Findability, Awareness, Market trust, Engagement) and the only place the offerings are named (¶4: AI visibility audits, recurring retainers, LLM content production). The body announces both lists (¶22, ¶35) with no items. Treat the pillar and offering names as reconstructed from the summary, not from the argument. **The argument, in order.** Trend (¶7 to ¶10): buyers ask AI platforms instead of scanning results; the labels are GEO and AEO. Commercial thesis (¶11 to ¶12): clients already ask why ChatGPT doesn't mention them, and agencies can sell the answer, ideally with SiteSonar. Mechanism (¶13 to ¶17): retrieval, synthesis, citation, and the four properties that get rewarded. Product (¶18 to ¶32): SEO tools measure the wrong things; SiteSonar tracks citations, finds competitor gaps, drafts pages, and publishes to WordPress behind an approval gate. Monetization (¶33 to ¶37): package it beside existing services; DoodleWeb's result. Close (¶38 to ¶41): GEO as retention, a layer on SEO. **What GEO is, separated from what SiteSonar claims.** The post's own description of the mechanism is sound as far as it goes. Retrieval-augmented answer engines fetch pages, pick sources, and cite them; a page that is fetchable, parseable, and answers a question in a form the model can lift is more likely to be cited than one that isn't. That's the same reasoning behind featured snippets and it predates this post. The term "Generative Engine Optimization" comes from Aggarwal et al., a Princeton and Georgia Tech paper first posted to arXiv in November 2023 and published at KDD 2024; the paper's benchmark found that adding citations, quotations, and statistics to content raised visibility in generative responses by up to 40% on their metric. The post does not cite that paper or any other measurement. The product layer sits on top of the mechanism and adds four claims: that SiteSonar can measure "prompt visibility rates" (¶25), that it can attribute the source material behind a competitor's citation (¶27), that it can draft gap-filling pages (¶29), and that it can publish them through a plugin behind an approval gate (¶31 to ¶32). None of those is demonstrated in the text. The one that is checkable from first principles is referral traffic from LLM interfaces (¶25): chatgpt.com, perplexity.ai, and gemini.google.com send referrer headers, so any analytics tool can count them. "Prompt visibility rate" is a vendor-defined metric whose denominator (which prompts, how many runs, which platforms, which dates) the post never states. **The measurement problem the post skips.** Generative answers are non-deterministic. The same prompt to the same platform on the same day returns different sources across runs, and platforms change retrieval behavior without notice. A "visibility rate" therefore has variance, and an 83% increase means nothing without the run count and the confidence interval. The post says the baseline "updates as AI answers change" (¶26), which concedes the instability and then presents it as a feature. **Every number, with its source.** There is one. ¶36: an 83% increase in measured AI visibility at DoodleWeb, plus unquantified citations and ChatGPT referrals, sourced to "SiteSonar's internal measurement." DoodleWeb is Sharma's agency and Sharma built SiteSonar (¶20): measurer, subject, and instrument vendor are one party. No window, no starting value, no traffic count. ¶7's "growing share" and ¶8's "no longer clicking" are trend assertions with no source. **The quote question.** The task brief describes a Chandan Sharma / DoodleWeb quote. The captured text attributes actions and motives to Sharma at ¶20 and ¶21 (he built SiteSonar after hearing from clients) and reports his agency's result at ¶36, but contains no sentence in quotation marks attributed to him. Any direct quote lives in the embedded video, which did not survive capture. Do not attribute quoted words to Sharma from this analysis. **Undisclosed interests.** WP Engine publishes the post. SiteSonar is a third-party product from an agency founder. The post does not state whether DoodleWeb is a WP Engine agency partner, whether SiteSonar integrates with WP Engine's platform, or whether any payment or referral arrangement exists. The absence of a disclosure is itself the finding: the relationship may be nothing more than a guest case study, but the reader can't tell. The registered-trademark footnote on "WordPress®" at ¶9 (link text "1") is the one place the WP Engine and Automattic trademark dispute shows through, and it appears as a footnote rather than a statement. **Where the reasoning holds.** ¶16 is hedged correctly, ¶17 keeps SEO in scope, ¶32's approval gate addresses a real failure mode, and ¶33 and ¶41 position GEO as adjacent to a current retainer rather than a new practice. Those are the parts a competitor can agree with in public; the Logical validity section takes each in turn. ## 🗒️ Notes #### Summary block (¶0 to ¶6) - Five compressed claims: discovery is shifting to conversational AI (¶0), GEO and AEO are a commercial expansion for agencies (¶1), FAME has four pillars (¶2), SiteSonar operationalizes GEO (¶3), agencies can productize into audits, retainers, and LLM content production (¶4). - ¶4 repeats the 83% DoodleWeb figure without the "internal measurement" qualifier that ¶36 carries. - Only ¶2 names the FAME pillars: Findability, Awareness, Market trust, Engagement. The body's list at ¶22 is empty in the capture. #### Introduction (¶7 to ¶12) - ¶7 compares the current change to mobile indexing taking over, with no date or figure for either. - ¶8 names four platforms: ChatGPT, Claude, Perplexity, Google AI Overviews. None is linked. - ¶11 is the demand thesis: clients are asking why ChatGPT doesn't mention them. ¶12 answers with SiteSonar plus "repeatable GEO service offerings." > [!quote] [¶11] > Clients are starting to ask a different kind of search question: “Why isn’t ChatGPT mentioning us?” #### How AI search is changing online discovery (¶13 to ¶17) - ¶13 to ¶14: ranked results ask the user to choose; generative engines synthesize one answer. - ¶15: the worked query is a WordPress agency in the Pacific Northwest specializing in higher-ed Drupal-to-WordPress migrations. That is an agency-shaped query, aimed at the reader, not at the reader's clients. - ¶16 is the mechanism sentence, correctly hedged with "may." - ¶17 lists the four rewarded properties. This is the vendor-neutral core of the post. > [!quote] [¶16] > the AI system may retrieve information from the web, evaluate relevant sources, synthesize an answer, and provide citations supporting its response. > [!quote] [¶17] > AI visibility also rewards clear entity relationships, structured information, broader brand authority, and content that directly answers the questions potential customers are asking. #### SiteSonar and AI search visibility (¶18 to ¶21) - ¶18 argues SEO platforms measure rankings, keywords, backlinks, and organic traffic, not LLM outputs and citations. Accurate as a description of 2023-era tooling; several SEO suites added AI-answer tracking in 2025, which the post doesn't mention. - ¶19 coins "Adaptive Optimization Engine" and promises "optimal rankings," an SEO word that doesn't describe how a synthesized answer cites sources. - ¶20: Chandan Sharma, founder of DoodleWeb, built SiteSonar. "As discussed in the video above" refers to media not in the capture. - ¶21: origin story (clients asked about ChatGPT) and scope (citation tracking, prompt visibility, content gap remediation). > [!quote] [¶18] > Traditional SEO platforms were built around search rankings, keywords, backlinks, and organic traffic rather than conversational LLM outputs and AI citations. > [!quote] [¶20] > Chandan Sharma, founder of the digital agency DoodleWeb, built SiteSonar specifically for agencies managing AI visibility across WordPress websites. #### The FAME framework for GEO (¶22 to ¶23) - ¶22 introduces four areas; the list items are absent from the capture. Names come from ¶2 only. - ¶23 states the goal: enough context and authority around a brand that AI systems surface it. This is a brand-level, not page-level, objective, which is the one conceptual shift in the post worth keeping. - Nothing ties any pillar to a specific technical control (schema type, page structure, link source). ¶2 mentions "Schema markup" once. > [!quote] [¶23] > the goal is enough context and authority around a brand that AI systems can confidently surface it when relevant. #### Improving AI visibility with SiteSonar (¶24) - One transition sentence; no new information. #### Citation tracking and visibility analytics (¶25 to ¶26) - ¶25: three tracked things: prompt visibility rates, citation placements, referral traffic from LLM interfaces. Only the third is measurable by anyone with server logs or analytics. - ¶26: the baseline "updates as AI answers change." That is a concession about non-determinism framed as a feature. #### Competitor blueprinting and gap analysis (¶27 to ¶28) - ¶27: analyze queries where a competitor is cited and examine the source material behind those citations. Plausible where platforms expose citations (Perplexity, AI Overviews, ChatGPT with browsing); not possible where they don't. - ¶28 turns "Why is the competitor showing up?" into a content gap work plan. The move from question to work plan is the sales mechanic of the post. > [!quote] [¶28] > this turns an otherwise abstract question, “Why is the competitor showing up?”, into an actionable content gap analysis with a clear work plan attached. #### The automated resource hub (¶29 to ¶30) - ¶29: the engine drafts "structured micro pages" with headings, data comparisons, and FAQs targeted at identified gaps. - ¶30: production is continuous, tied to prompts where visibility is available, "and keeps producing as new gaps appear." That is the scaled-content pattern; the post's only guard is the approval gate two paragraphs later. #### Native WordPress integration and governance (¶31 to ¶32) - ¶31: review, edit, push to WordPress via a dedicated plugin. The plugin's name, its WordPress.org listing status, and its permission model are not given. - ¶32: nothing goes live without explicit agency or client approval. This is the post's strongest governance statement and the one to hold the vendor to. > [!quote] [¶32] > No content goes live without explicit agency or client approval #### Monetizing GEO as an agency service (¶33 to ¶37) - ¶33: GEO sits beside SEO, content strategy, WordPress development, and ongoing optimization. ¶34: package as recurring or premium add-on. - ¶35 promises four offerings; the list is absent. ¶4 names three: AI visibility audits, recurring retainers, LLM content production. - ¶36 is the only statistic in the post and its provenance is the vendor's own tool on the vendor's own agency. > [!quote] [¶36] > According to SiteSonar’s internal measurement, DoodleWeb recorded an 83% increase in its measured AI visibility #### GEO as a client retention strategy (¶38 to ¶41) - ¶38: organic visibility now spans search engines, AI answers, and conversational platforms. - ¶39: the agency's expanded job is to show clients where they appear, what sources AI cites, where competitors are stronger, and what content closes gaps. That's a reporting product; the post never prices it. - ¶41 closes with the layer-not-replacement framing. > [!quote] [¶41] > GEO doesn’t need to replace SEO. ## 📚 References The article contains 8 outbound anchors. All are listed below with the claim each one supports, followed by primary sources for named entities that the post does not link. - [Technical GEO and AI Search Guide (WP Engine)](https://wpengine.com/blog/technical-geo-ai-search/: linked twice, from ¶5 ("Review the Technical GEO...") and from the phrase "Generative Engine Optimization" at ¶10. It is where the post sends readers for the mechanics; this post itself contains no technical detail beyond ¶16 and ¶17. - [WordPress® footnote, anchor 1](https://wpengine.com/blog/geo-for-agencies/#00afe078-7009-4914-8a3b-a4db977cc96b) and its [return link ↩︎](https://wpengine.com/blog/geo-for-agencies/#00afe078-7009-4914-8a3b-a4db977cc96b-link: the registered-trademark footnote attached to "WordPress" at ¶9. Supports no claim; it is a legal notice. - [Chandan Sharma on LinkedIn](https://www.linkedin.com/in/chandansharma16/: identifies the SiteSonar founder named at ¶20. - [DoodleWeb](https://doodleweb.io/: the agency at ¶20 and the case-study subject at ¶36. The link goes to the agency homepage, not to any published result. - [SiteSonar](https://sitesonar.ai/: linked twice, at ¶12 and ¶19. Vendor homepage. None of the capabilities at ¶25 to ¶32 or the 83% at ¶36 is supported by a linked document; the link is to the product, not to evidence. Primary sources for entities the post names but does not link: - [GEO: Generative Engine Optimization, Aggarwal et al., arXiv 2311.09735 (KDD 2024)](https://arxiv.org/abs/2311.09735: origin of the term at ¶10. The post does not cite it. It is the only published benchmark of content changes against generative-engine visibility and the natural counterweight to the vendor-defined metric at ¶25. - [Google Search Central: AI features and your website](https://developers.google.com/search/docs/appearance/ai-features: Google's own statement of what AI Overviews use, relevant to the platform named at ¶8 and to the claim at ¶17. - [Google Search spam policies: scaled content abuse](https://developers.google.com/search/docs/essentials/spam-policies#scaled-content-abuse: the policy that bears on continuous AI-drafted micro pages at ¶29 to ¶30. - [OpenAI: GPTBot and crawler documentation](https://platform.openai.com/docs/bots: how ChatGPT retrieval reaches a site, the precondition for any citation claimed at ¶25. - [Schema.org](https://schema.org/: the "Schema markup" named once at ¶2 and implied by "structured information" at ¶17. ## 🗣️ Paraphrase WP Engine's pitch to agencies runs like this. Your clients' buyers now ask ChatGPT, Claude, Perplexity, and Google's AI Overviews for recommendations instead of scanning ten results, and those systems fetch pages, pick sources, and cite them inside a single answer. A client site gets cited when its entities are clear, its information is structured, its brand carries authority beyond its own domain, and its pages answer the exact question a buyer types. Your clients have noticed; they're asking why ChatGPT doesn't mention them. That question is a service you can sell. The post then hands you a tool to sell it with. SiteSonar, built by agency founder Chandan Sharma, watches how AI platforms cite a brand, finds the prompts where a competitor gets cited instead, drafts pages to fill those gaps, and pushes them to WordPress through a plugin, with a human approving each one before it goes live. The FAME framework (Findability, Awareness, Market trust, Engagement) names the four areas the work covers. Package it as an audit, a retainer, or content production, and it sits next to the SEO work you already bill for. Sharma's own agency saw its AI visibility rise 83%, measured by SiteSonar. GEO doesn't replace SEO; it's one more layer of keeping clients visible wherever people search. What it doesn't hand you: a measurement from outside the vendor, a definition of the metric, WP Engine's relationship to the tool, or the two lists it announces. ## 📰 Newsletter **WP Engine just told its agencies what to sell next.** Brian Gardner's post on GEO for agencies is a full go-to-market kit in 1,700 words: a trend (buyers ask ChatGPT instead of Googling), a framework (FAME: Findability, Awareness, Market trust, Engagement), a tool (SiteSonar, built by DoodleWeb's Chandan Sharma), a packaging section (audits, retainers, content production), and a case study (83% more AI visibility at DoodleWeb). If you run or support WordPress agencies, read it as competitive intelligence about how a host arms its channel. The keepable part is vendor-neutral and costs nothing: AI answer engines cite pages that are fetchable, structured, clearly about a named entity, and that answer the question directly. That's ¶17 of the post and it's also the 2023 Princeton GEO paper, which the post doesn't cite. Everything specific sits on the product side, and the one number in the piece is SiteSonar measuring SiteSonar's founder's agency with SiteSonar's own metric. Carry the mechanism into your own advice. Leave the number where you found it. One thing to check before anyone sells GEO on your platform: whether your CDN or WAF blocks GPTBot, ClaudeBot, or PerplexityBot by default. No crawl, no citation, no matter how good the content. Links: [the post](https://wpengine.com/blog/geo-for-agencies/), [WP Engine's technical GEO guide](https://wpengine.com/blog/technical-geo-ai-search/), [the GEO paper](https://arxiv.org/abs/2311.09735), [Google on AI features](https://developers.google.com/search/docs/appearance/ai-features). ## ✅ Action items #### DevRel - Developer advocacy: build and publish an open measurement script that runs a fixed prompt set (20 prompts) against three AI platforms 10 times each, records which domains get cited, and reports the citation rate with its variance. A reproducible baseline is the counter to a vendor-defined "prompt visibility rate," and it's a talk in itself. - Education: run one office-hours session for agency partners titled "GEO without buying anything": schema.org JSON-LD in WordPress, entity pages, FAQ structure, and how to read LLM referral traffic in analytics. Use ¶17 as the outline. - Developer experience: publish a one-page "is your WordPress site citable" checklist: robots.txt allows the named AI crawlers, pages render server-side, JSON-LD validates, one canonical entity page exists for the organization. Examples first, no product. - Developer marketing: track this post as a competitor move. WP Engine now has framework plus tool plus packaging for the agency channel. Log whether Pantheon has an equivalent narrative for agency partners and, if not, whether it should. - Community management: when the 83% figure circulates in community channels, attach the ¶36 provenance in one line ("vendor's own measurement of the vendor's own agency") rather than disputing it. - Internal/external/cross-project: ask WP Engine's DevRel counterparts directly whether SiteSonar is a partner integration. A one-line answer settles the disclosure question this post leaves open. #### OSPO - Policy & Governance: decide whether AI-drafted content pushed through a third-party plugin is in scope for any open-source contribution guidelines the organization maintains. If a contributor uses such a tool on project docs, the approval gate at ¶32 is the minimum bar. - License Compliance: SiteSonar's terms on the content it drafts (ownership, training on client content) are not stated in the post. Add "who owns AI-drafted pages" to the vendor review checklist. - Security Management: a plugin that publishes content from an external service holds publish capability on client sites. Treat it as a supply-chain dependency and ask for its permission model before recommending it. - Software Inventory: none from this source. - Developer Support: none from this source. - Community Engagement: none from this source; nearest relevance is whether any WordPress project team wants a shared position on AI crawler access for .org properties. - Strategic Alignment: no direct action; nearest relevance is a KPI for LLM referral traffic to owned properties, which is measurable today from referrer headers. #### Community - Maintainers: none from this source. - Contributors: none from this source. - Sponsors: if a sponsor funds a GEO pilot on a community site, require the prompt set, run count, and dates be published with any before/after number. - Ecosystem: none from this source; nearest relevance is a cross-project comparison of how CMS documentation sites get cited by AI answer engines. #### WordPress - Core: no action. Nothing here touches core code. - Core Performance: none directly. AI crawlers add request volume to origin; a site that serves crawlers uncached pays for citation in TTFB. Worth one line in performance guidance. - Core AI: if the Core AI team publishes guidance on AI-generated content in WordPress, ¶32's human approval gate is the pattern to cite, and continuous generation at ¶30 is the anti-pattern. - **Core Program Team:** none from this source. - Mobile: none. - Playground: a Playground blueprint that ships a site with valid JSON-LD, an organization entity page, and an FAQ block would let anyone test citability claims in a browser. Low cost, high teaching value. - Community: none from this source. - Training: draft a Learn WordPress lesson on structured data and entity pages framed for AI answer engines. ¶17 supplies the four learning objectives. - Marketing: do not repeat the 83% figure in any WordPress or Pantheon material without "SiteSonar's internal measurement of its founder's agency" attached. Separately, decide within 30 days whether AI search visibility is a Pantheon agency-partner narrative; measure LLM referral traffic to pantheon.io by referrer (chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com) as the baseline before deciding. - Support: expect tickets asking "why doesn't ChatGPT show my site" (¶11 says clients already ask). Prepare one knowledge-base answer that covers crawler access, structured data, and the fact that answers vary run to run. - Design: none. - Documentation: write one docs page for the platform: which AI crawlers the default configuration allows or blocks, how to change it, how to verify a page is fetchable by GPTBot and ClaudeBot, and how to count LLM referral traffic. Examples-first, with the exact user-agent strings. This page doesn't exist for most hosts and it's the first thing an agency selling GEO will need. - Photos: none. - TV: none. - Polyglots: none. - Accessibility: none from this source. - Meta: none from this source; nearest relevance is whether WordPress.org's own robots.txt allows AI crawlers. - CLI: none. - Hosting: audit the platform's default bot handling within two weeks. If the CDN, WAF, or rate limiter blocks or challenges GPTBot, ClaudeBot, PerplexityBot, or Google-Extended by default, every GEO promise made on the platform fails at the crawl step. Publish the default, make it configurable per site, and log AI-crawler hits separately from AI-referral visits so agencies can report both. - Tide: none. - Openverse: none. - Themes: none from this source; nearest relevance is whether default block themes emit JSON-LD for the site entity. - Plugins: the post's "dedicated WordPress plugin" at ¶31 is unnamed. If it appears on WordPress.org, its review should cover remote-publish capability and the approval gate ¶32 promises. #### The WPOCC - Programs: none from this source. - Projects: none from this source; nearest relevance for The WP Open Community Collective is a documentation sprint on AI crawler access for community-run sites. - Funding experiments: none from this source. ### FAIR - Technical independence: AI answer engines are a new external dependency for discovery, and none of the four platforms named at ¶8 publishes a stable citation contract. Note it as an independence risk alongside the existing ones. - Policies and covenants: none from this source. - Productized governance artifacts: none from this source; nearest relevance is a scorecard row for "AI crawler policy published" on operator sites. ## 🧪 Hypotheses 1. Referral traffic with LLM-interface referrers (chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com) is below 2% of organic sessions for the median WordPress agency client site in 2026. ¶7's "growing share" is testable and probably small in absolute terms. 2. For a fixed prompt set run 10 times per platform, the standard deviation of a domain's citation rate exceeds 15 percentage points, which makes a single before/after percentage like ¶36's 83% uninterpretable without the run count. 3. Sites blocked to GPTBot or ClaudeBot at the CDN level receive zero citations from the corresponding platform regardless of content quality, so crawler access predicts citation better than any FAME pillar. 4. Adding valid schema.org JSON-LD for Organization and FAQPage to a page raises its citation rate in AI Overviews and Perplexity by a measurable amount within 60 days; the effect on ChatGPT is smaller or absent. ¶17's "structured information" is testable per platform. 5. Continuous AI-drafted micro pages (¶29 to ¶30) published at more than 20 per month on a single site trigger a measurable Google organic decline within two quarters, meaning the GEO tactic and the SEO tactic conflict on the same site. 6. Agencies that add a GEO reporting line (¶39) to an existing SEO retainer see lower churn than those that don't, independent of any change in client visibility, because the report itself is the retention mechanism. ## 📊 Methodology - **H1.** 12 months of analytics from 30 consenting client sites; classify sessions by referrer host; report LLM-referrer share of organic sessions as a distribution, not a mean. - **H2.** 20 fixed prompts, 10 runs each on ChatGPT, Perplexity, and AI Overviews on one day; record cited domains; compute citation rate and standard deviation per domain; repeat weekly for 8 weeks to capture drift. - **H3.** 40 sites, 20 blocked to AI crawlers via robots.txt and CDN rules, 20 open, content held constant; run the H2 prompt set at weeks 0 and 8; confirm blocks from server logs by user agent. - **H4.** 20 sites, Organization and FAQPage JSON-LD added to treatment pages with matched controls; validate markup; run the H2 prompt set at weeks 0, 4, and 8; report effect sizes per platform. - **H5.** 15 sites using continuous AI content production matched to 15 with editorial cadence; two quarters of Search Console impressions and clicks; observational, so state the confounders. - **H6.** Survey 50 agencies on 12-month retainer churn, split by whether they added an AI-visibility report; pair with H2 measurements to separate reporting from results. - **Controls throughout.** Publish the prompt set before running it. Log platform versions and dates. Report variance with every rate. Never use a vendor's own dashboard as the measurement instrument for that vendor's claim. ## ❓ Questions - What is the denominator of "prompt visibility rate" at ¶25: how many prompts, how many runs, which platforms, over what dates? - What was DoodleWeb's starting value before the 83% increase at ¶36, and over what period? - Does WP Engine have a commercial, partner, or integration relationship with SiteSonar or DoodleWeb? The post doesn't say. - What is the WordPress plugin at ¶31 called, is it on WordPress.org, and what capability does it require on the client site? - Which of the four platforms at ¶8 expose citations in a form a tool can parse, and which does SiteSonar actually monitor? - Who owns the micro pages SiteSonar drafts at ¶29, and does the vendor train on client content? - What are the four FAME pillars in the author's own words, and the four offerings at ¶35? Both lists are absent from the capture. - How does continuous page generation at ¶30 square with Google's scaled content abuse policy on the same site? ## 🔮 Future research - **Citation determinism across platforms.** No public dataset measures how stable AI-answer citations are for the same prompt over time. Without that, every visibility metric is noise-bounded by an unknown. - **Crawler access as the first-order variable.** Hosts control robots.txt defaults and bot rules for millions of sites. Whether default crawler policy explains more citation variance than content changes is unstudied and directly actionable by hosts. - **GEO and SEO conflict on one domain.** The post treats them as layers (¶41). If scaled AI-drafted pages depress Google organic while raising AI citations, agencies are trading one channel for another and should know the exchange rate. - **Vendor-defined metrics as a category.** "Prompt visibility rate" joins "domain authority" and "share of voice" as proprietary numbers sold back to the people they measure. A taxonomy of which such metrics are reproducible would help agencies price tooling. - **The disclosure norm for host blogs.** Whether readers can tell a paid placement from a case study, and whether it changes purchase decisions, is measurable and unmeasured. ## 💡 Implications - **Practice.** Agencies can start GEO with zero tooling: allow AI crawlers, add valid JSON-LD, build one entity page per client, and count LLM referrers. Everything in ¶17 is doable in a week. The tool decision can wait for a baseline. - **Policy.** Hosts that block AI crawlers by default are making a GEO decision for every customer. That default should be published and configurable, and the choice belongs to the site owner, not the platform. - **Security.** A plugin that accepts content from an external service and publishes it holds a publish-capable credential on the client site. Approval gates (¶32) reduce content risk but not credential risk. Vendor review should ask how the plugin authenticates and what it can do if the vendor is compromised. - **Education.** The mechanism at ¶16 and ¶17 is teachable in one lesson and maps onto structured-data curriculum that already exists. Teach the mechanism with the Princeton paper as the source, not with a vendor post. - **Funding.** GEO retainers are priced on a metric the vendor defines and the vendor measures. Clients paying for "visibility" should ask for the prompt set and the run count, and agencies should be able to reproduce the number without the tool. - **Competitive.** WP Engine is arming its agency channel with a complete narrative. A competing host without an AI-search story for agencies will hear about this post from its own partners. ## 🧠 Learning styles assessment - **Reading level estimate.** Flesch-Kincaid Grade Level on the 1,101-word body (¶7 to ¶41): 14.9; Flesch Reading Ease 23.8. Sentences average 20 words at 1.92 syllables per word. The density comes from compound noun phrases ("Adaptive Optimization Engine for GEO and AEO," "content gap remediation"), not sentence length: whitepaper register for agency owners. #### 👀 Visual learners - Present: an embedded video (¶20); the two lists that vanished in capture were likely visual cards. - Improve: no diagram of the ¶40 loop (track, analyze, draft, publish, approve). If the pillars are an image, the capture proves they have no text fallback. #### 🔊 Aural learners - Present: the embedded video, contents unknown from the capture. - Improve: a 3-minute audio summary of ¶16 and ¶17; verify the video's captions. #### 🗣️ Verbal learners - Present: GEO and AEO expanded at ¶10; the mechanism in plain sentences at ¶13 to ¶17; the ¶11 client question as a hook. - Improve: "Adaptive Optimization Engine," "prompt visibility rate," and "citation placements" go undefined, and FAME is named without explanation in the body. One sentence each. #### 🧩 Kinesthetic learners - Present: none; no step to try, no query to run. - Improve: a five-step "test your own site" box (three prompts on Perplexity, check citations, check robots.txt for GPTBot, validate JSON-LD, check referrers). #### 🧠 Logical learners - Present: clean structure (trend, mechanism, tooling gap, tool, packaging, result, close); the ¶16 hedge is correctly placed. - Improve: the one statistic has no baseline or method. Show the starting value, prompt set size, and period. #### 🧑‍🤝‍🧑 Social learners - Present: written for agencies to take to clients; ¶11 is a conversation opener. - Improve: no community or forum link, and WP Engine's agency programs go unmentioned. #### 🧘‍♀️ Solitary learners - Present: an 8-minute read with a linked next step (the Technical GEO guide). - Improve: no reflection prompt. "Which three questions do your clients' buyers ask an AI, and would the site answer them?" #### 🌿 Naturalistic learners - Present: nothing explicit. - Improve: continuous generation and AI crawling both cost compute and origin load; an energy-per-citation framing would land here. ## 📊 Accessibility assessment Judgments are about the page as captured, not a live audit; the capture kept headings, paragraphs, and anchors and lost two lists and the video. #### Screen reader semantics - Present: 10 headings (H2 to H4) that follow the argument; real paragraphs. - Improve: two announced lists (¶22, ¶35) have no text in the capture, which suggests images or scripted components without text equivalents: 1.1.1 (Non-text Content) if confirmed. Footnote link text "1" and return link "↩︎" are non-descriptive: 2.4.4 (Link Purpose, In Context). #### Keyboard navigation - Present: a standard WordPress blog layout is keyboard operable by default. - Improve: the embedded video player and any card carousel for the pillars need focus order and visible focus checked. 2.1.1 (Keyboard) and 2.4.7 (Focus Visible) unverified from the capture. #### Contrast and readability - Present: unverifiable from text. - Improve: WP Engine's brand palette uses light gray body text on white in some templates; check 1.4.3 (Contrast Minimum) at 4.5:1 on body copy and on the summary block. #### Text resizing and reflow - Present: WordPress themes reflow by default. - Improve: if the FAME pillars are a fixed-width graphic, text inside it doesn't resize. 1.4.4 (Resize Text) and 1.4.10 (Reflow) at risk for that element. #### Media and captions - Present: an embedded video (¶20). - Improve: the post relies on the video for Sharma's own words. Without verified captions and a transcript, that content is unavailable to deaf readers and to text-only capture, which is what happened here. 1.2.2 (Captions, Prerecorded) and 1.2.3 (Audio Description or Media Alternative) to verify. #### Forms and interactions - Present: none in the article body. - Improve: the page likely carries a newsletter or demo CTA form outside the captured body. 3.3.2 (Labels or Instructions) applies there. #### Standards reference - Primary risks against WCAG 2.2 AA: **1.1.1** (lists that vanished in text capture), **2.4.4** (footnote link text "1" and "↩︎"), **1.2.2** (video captions unverified, and the video carries content the text lacks), **1.4.4 / 1.4.10** (graphic pillar text). The two missing lists are the highest-impact item because they hold the framework and the offerings the post is selling. ## 🌍 DEIB assessment #### Cultural breadth - Present: the four platforms named are global services. - Improve: every locality is American: the Pacific Northwest (¶15), "major metropolitan markets" (¶36), US higher-ed Drupal migrations. Non-English AI answer behavior differs and is never mentioned. #### Inclusive language - Present: gender-neutral throughout; "clients," "buyers," "agencies." Chandan Sharma is referred to by name, without pronouns. - Improve: "ten blue links" (¶8) is search-industry idiom that assumes insider knowledge. GEO, AEO, LLM, and Schema are used before or without definition. #### Neurodiversity support - Present: consistent structure, one idea per section, headings every two or three paragraphs. - Improve: the summary block and the body repeat the same claims in different words, which costs re-reading. Undefined coinages ("Adaptive Optimization Engine") force the reader to infer meaning from context. #### Representation - Present: one named person, an agency founder of South Asian name, presented as builder and expert. - Improve: he is also the only case study and the tool's vendor, so the single voice is also the interested party. No client, no independent practitioner, no platform representative appears. #### Psychological safety - Present: the tone is measured; competitors are "cited instead of your client," not attacked. - Improve: the hook at ¶11 ("Why isn't ChatGPT mentioning us?") and the retention framing at ¶38 to ¶41 lean on fear of losing clients. Stating what share of discovery actually runs through AI today would lower the pressure and let agencies size the opportunity honestly. ## ✅ Logical validity #### Strong reasoning - **Conditional hedge, ¶16.** "Depending on the platform and query, the AI system may retrieve..." is the correct modality for a behavior that varies by platform and changes over time. It's the most carefully worded sentence in the post. - **Retained premise, ¶17 and ¶41.** Keeping SEO fundamentals in scope, and closing with "GEO doesn't need to replace SEO," avoids the false dilemma most AI-search marketing falls into. The argument is additive, which matches how agencies actually add services (¶33). - **Governance as design, ¶32.** Requiring explicit approval before publish addresses the real failure mode of automated content, and it's stated as a hard rule ("No content goes live without"), not a preference. Strengthen by naming who can approve and whether approval is logged. - **Question to work plan, ¶28.** Converting "Why is the competitor showing up?" into a content gap analysis is a valid problem-decomposition move and it's the sales mechanic that makes the service sellable. Strengthen by showing one worked example with the competitor's cited source. - **Adjacent-service reasoning, ¶33.** GEO sits beside SEO, content strategy, and WordPress development, so the marginal cost of offering it is low. That's an argument from existing capability, and it's sound as far as it goes. #### Fallacies - **Unsupported trend assertion, ¶7 and ¶8.** "a growing share of discovery now happens inside AI platforms" and "Buyers are no longer clicking through ten blue links." No figure, no source, no date. The second overgeneralizes: buyers still click; the share has shifted by an unstated amount. *Fix:* cite one measured share of referral traffic from AI interfaces, with the date and the population. - **Self-referential evidence (conflict of interest), ¶36.** The tool measured its own vendor's agency with its own metric and reported an 83% gain. The instrument, the subject, and the interested party are the same. *Fix:* an independent measurement, or at minimum the prompt set, run count, baseline, and dates so a reader can reproduce it. - **Hasty generalization, ¶37.** One self-measured agency becomes "a useful example of how GEO can move beyond experimentation and become a measurable agency service." One case, and the case is the founder's. *Fix:* three client agencies not affiliated with the vendor. - **Persuasive definition (reification), ¶19.** "Adaptive Optimization Engine" is capitalized as if it names a category; it names one product. The same sentence promises "optimal rankings," a ranked-list concept that doesn't apply to a synthesized answer. *Fix:* lowercase verbs, and "citation" in place of "rankings." - **Suppressed information, ¶20 and throughout.** The post never states WP Engine's relationship to SiteSonar or DoodleWeb, and never names the WordPress plugin at ¶31. The reader can't evaluate the recommendation's independence. *Fix:* a one-line disclosure at the top, and the plugin's name. #### Tone of voice - Measured B2B marketing register in the third person, with the reader addressed as "agencies" and "you" only implicitly. Hedged verbs at ¶16 ("may retrieve") and ¶33 ("doesn't require") sit beside confident product verbs at ¶19 ("continuously monitors," "builds and adjusts"). Chandan Sharma becomes "Chandan" on second reference (¶21, ¶36), a first-name familiarity that signals a partner relationship without stating one. There are no superlatives and no exclamation marks; the persuasion runs through structure (trend, gap, tool, result) rather than through adjectives. The place to read carefully is ¶36, where the only number arrives wrapped in the most qualifiers in the post ("According to SiteSonar's internal measurement," "its measured AI visibility"). The author knew the provenance was weak and said so, quietly. ## 🛡️ DARVO analysis Not applicable to this source. DARVO (Deny, Attack, Reverse Victim and Offender) describes a response to accusation or criticism. This is a vendor go-to-market post with no accuser and no defensive posture. For completeness across the three prongs: **Deny**: nothing is denied. The one place a live dispute touches the text is the registered-trademark footnote on "WordPress®" at ¶9, which is a legal notice, not a denial or a minimization. **Attack**: no party is criticized. Competitors appear only as sites "cited instead of your client" (¶27), and traditional SEO tools are described as built for a different job (¶18), not as failures. **Reverse victim/offender**: no such move. The nearest emotional lever is the client-loss framing at ¶11 and ¶38 to ¶41, which is ordinary fear-of-missing-out marketing rather than a victim reversal. ## 🎯 Conclusion Strip the product away and the post says something true and cheap to act on: answer engines fetch, select, synthesize, and cite, and they favor pages with clear entities, structured data, offsite authority, and direct answers (¶16, ¶17). That's the 2023 GEO paper restated for agency owners, and it requires no purchase. The post also gets the shape of the sale right: the client asks "Why isn't ChatGPT mentioning us?" (¶11), and an agency that can answer with a report and a work plan has a new retainer line beside the SEO one it already bills (¶33, ¶39). Everything specific is unverified in this source. The only number, 83%, is the vendor measuring its own founder's agency with its own metric (¶36), with no baseline, period, or prompt set. The pillars and the offerings are absent where the body announces them. WP Engine's relationship to SiteSonar is not stated, the plugin is not named, and the non-determinism that makes any "visibility rate" hard to trust is mentioned once (¶26) as a feature. **Next steps for readers.** If you advise agencies: teach the mechanism from ¶17 and the Princeton paper, and give them the free checklist first (crawler access, JSON-LD, entity page, LLM referrers). If you work at a host: audit what your defaults do to GPTBot, ClaudeBot, PerplexityBot, and Google-Extended this month, because no content strategy survives a blocked crawl, and write the docs page that says so. If you work in marketing: file this post as WP Engine's agency-channel play for AI search, and decide whether you need one. In every case, carry the 83% nowhere without the sentence it came wrapped in. ## 🔄 Refactor Appendix — GEO for agencies: Monetizing AI search visibility > Density: 2–5 Literature atoms (H3 pad), 0–3 Permanent candidates (H4 pad). > Emitted: 3 Literature, 1 Permanent. Resonance filter applied: the FAME pillar > names were dropped as atoms because they exist only in the summary block, and the > capability claims at ¶25 to ¶31 were dropped as undemonstrated. > > Konik claim-header rules: present tense, declarative, ≤10 words, searchable, unique. > > Atomicity: one declarative sentence per atom. "and"/"also" → split. > > Tag pool: #bmf · #addendum · #xref · #articleseed · #storystem · #share · #aphorisms ### 📚 Literature Split Pad #### [H1] AI answers reward entities, structure, authority, direct answers > [!quote] [¶17] > AI visibility also rewards clear entity relationships, structured information, broader brand authority, and content that directly answers the questions potential customers are asking. **Context & use**: - ¶17, the last sentence of the mechanism section, immediately after the hedged description of retrieve-evaluate-synthesize-cite at ¶16. - This is the vendor-neutral core of the post and the only sentence a competitor can teach verbatim. It maps one-to-one onto a free checklist: entity page, JSON-LD, offsite mentions, FAQ-shaped content. The Princeton GEO paper (arXiv 2311.09735) is the citation the post should have used. **Tags**: #xref #share  #### [H2] The post's only statistic is vendor self-measurement > [!quote] [¶36] > According to SiteSonar’s internal measurement, DoodleWeb recorded an 83% increase in its measured AI visibility **Context & use**: - ¶36, the case-study paragraph in the monetization section. DoodleWeb is the SiteSonar founder's own agency (¶20). - Keep as a specimen: the author attaches two qualifiers ("internal measurement," "its measured") that signal the weakness without stating it. No baseline, period, prompt set, or run count. Any reuse of 83% must carry this sentence. **Tags**: #xref #share #articleseed  #### [H3] A recurring client question names the new service line > [!quote] [¶11] > Clients are starting to ask a different kind of search question: “Why isn’t ChatGPT mentioning us?” **Context & use**: - ¶11, the pivot from trend to commercial thesis. ¶28 repeats the move with the competitor version of the question and turns it into a work plan. - The go-to-market pattern is reusable well beyond GEO: the demand signal arrives as a question clients already ask, and the service is the answer packaged with a report. For DevRel, the questions agencies field are the cheapest product research available. **Tags**: #xref #articleseed  ### 🧠 Atomic Split Pad #### [P1] A client's repeated question sizes a service before any tool does **Claim** (one sentence, Courtney's voice): A question clients keep asking is a more reliable signal for a new service line than any vendor's market-size claim. **Evidence pointer**: [[#H3]] · [[#H2]] as the contrast case (the vendor supplies the number, the client supplied the question) **Scope & conditions**: Holds when the question recurs across unrelated clients and names an outcome they'd pay for ("mention us"). Breaks when the question is prompted by the vendor's own marketing, when only one client asks it, or when the honest answer is "nothing you can do," in which case the service is a report with no lever. The follow-on test is whether the agency can answer the question without buying the tool that surfaced it. **Links**: - [[Agency go-to-market]] — because extends - [[A precise number without a method is a marketing claim]] — because contrasts on where the signal originates (client question versus vendor metric) - [[Developer relations as product research]] — because applies-in DevRel intake from partner agencies  ### ✏️ Fleeting Queue - [F1] Confirm the four FAME pillar definitions and the four productized offerings from the live page or the WP Engine Technical GEO guide; both lists are absent from the capture at ¶22 and ¶35, https://wpengine.com/blog/geo-for-agencies/ #fleeting #inbox - [F2] Find out whether WP Engine has a partner, integration, or commercial relationship with SiteSonar or DoodleWeb; the post does not disclose one #fleeting #inbox - [F3] Identify the "dedicated WordPress plugin" at ¶31: name, WordPress.org listing, capability it requires on the client site #fleeting #inbox - [F4] Open question: on the platform I work with, do default CDN or WAF rules block or challenge GPTBot, ClaudeBot, PerplexityBot, or Google-Extended? No crawl means no citation regardless of content #fleeting #inbox - [F5] Article seed: "GEO without buying anything," a free checklist built from ¶17 and the Princeton GEO paper, aimed at agency partners who will read WP Engine's post this week #fleeting #inbox #articleseed - [F6] Measure LLM-referrer sessions (chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com) on one owned site for 30 days to put a real number under ¶7's "growing share" #fleeting #inbox ### → Things 3 (max 2 actions) - [ ] Audit the platform's default bot rules for GPTBot, ClaudeBot, PerplexityBot, and Google-Extended, and note whether each is allowed, blocked, or challenged — hosting docs gap, 30 min #things - [ ] Draft the "is your WordPress site citable" one-page checklist from ¶17 for agency partners, with no product named — DevRel content, 45 min #things  📄 **Full text:** [[Full Document Contents/Articles/GEO for Agencies Monetizing AI Search Visibility.md|GEO for Agencies: Monetizing AI Search Visibility by Brian Gardner]] ## Highlights - For agencies, the transition opens a new commercial opportunity. Clients are starting to ask a different kind of search question: “Why isn’t ChatGPT mentioning us?” That question points to real demand: helping clients gain visibility where more discovery and decision-making now happen. ([View Highlight](https://read.readwise.io/read/01m388pxdsw8fyj7hkqrgwgy0b)) - Note: [H3] A recurring client question names the new service line - Tags: atom/literature/h3 - That changes what agencies need to consider when optimizing client websites. Traditional SEO fundamentals remain important, but AI visibility also rewards clear entity relationships, structured information, broader brand authority, and content that directly answers the questions potential customers are asking. ([View Highlight](https://read.readwise.io/read/01m388ps0exjwtz01begdfr4th)) - Note: [H1] AI answers reward entities, structure, authority, direct answers - Tags: atom/literature/h1 - Chandan has applied this strategy within his own agency, DoodleWeb. According to SiteSonar’s internal measurement, DoodleWeb recorded an 83% increase in its measured AI visibility, secured citations for competitive web development queries in major metropolitan markets, and generated referral traffic from ChatGPT. ([View Highlight](https://read.readwise.io/read/01m388pv8nmsexp8epm4a6zrbg)) - Note: [H2] The post's only statistic is vendor self-measurement - Tags: atom/literature/h2