Is WordPress Really Losing Ground to AI?  🔗 **Source:** [Is WordPress Really Losing Ground to AI?](https://www.youtube.com/watch?v=PD5DYYJRmS0) > [!note]+ Document Notes > tags: [wordpress-ecosystem, wordpress-market-share, ai-site-builders, agency-business, web-performance] > author: WP Minute (Matt Medeiros with Remkus de Vries) > source: WP Minute Podcasts (YouTube) > reference: https://www.youtube.com/watch?v=PD5DYYJRmS0 > date created: 2026-09-21 > > > [!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 Do not quote or attribute any other line to the named author.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: Is WordPress Really Losing Ground to AI? > <!-- deep-read:sections start --> ## 🧭 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) · [🧰 Template Library](#-template-library) ## 🔗 References (in-page) - Source video: https://www.youtube.com/watch?v=PD5DYYJRmS0 — WP Minute Podcasts, uploaded 2026-09-21, 40m23s - Host: Matt Medeiros, The WP Minute's founder and usual host (not named in the captions). Guest: Remkus de Vries of withinwp.com; auto-captions render his name "Remiss." Both identifications rest on the show and the guest's domain, not on on-screen self-introduction. - **Recording date:** the host says "It is now August 4th" at [00:22:08], so the conversation was recorded about seven weeks before upload. Anything described as current is current as of early August 2026. - Caption spellings for other names mentioned (Mark "Zamansky," James "Welbs," Andy "Peting," "Mdash from Cloudflare") are unverified; check them before citing. ## 📝 Summary **What happened.** Matt Medeiros asks Remkus de Vries whether the wave of developers leaving WordPress for AI site builders, headless CMSs, and static generators reflects a better product or people chasing something shiny. De Vries argues it's neither simple story: the market-share numbers everyone argues about have never been solid, the defections come mostly from the bottom of the market, and the agencies and plugin businesses that are struggling mostly never had to differentiate themselves before. **The main arguments.** - **The data was never solid.** Market-share and WordPress-economy figures "have always been an assumption"; detection gets harder as sites strip the generator tag, rewrite plugin paths through optimizers, or publish static output. The declines people cite are incremental, and other CMSs aren't gaining much either. - **The bottom of the market is moving.** Low-budget site owners will increasingly let AI generate their sites. That market "was never going to spend much time in the WordPress economy anyway." The moment a site needs a second editor, complexity returns. - **Projects are moving upmarket.** De Vries knows agencies that struggle to take on anything under $50K because their discovery process alone would exceed it. - **Differentiation is the real crisis.** Since about 2010, agencies could win work just by knowing how to build; now they have to explain why a client should hire them. - **WordPress's AI capability is arriving.** De Vries points to 7.0's native AI interaction and expects the ecosystem to be "really mature" in connecting AI by the end of next year. He names Miles (a beta WordPress site-building tool he's used on five sites), WPVibe, and WP God Mode. - **Judging standards versus products.** Asked how to tell a lasting standard from a Google AMP-style fad, he says: is a big company promoting agreed protocols, or its own library? He counts Cloudflare's "is my site agent ready" checker as the former, and recommends Cloudflare's APO edge cache ($5 a month). - **His business.** The Within WordPress Guild, announced May 4 2026, aims to digest each WordPress release into about 30 pieces of principles-first training for agencies. He's explicit that he wants to be paid for it. **Why it matters.** It's a sober counterweight to "WordPress is dying" content, from someone who teaches performance for a living, and it puts the pressure on agencies' positioning rather than on the software. **Risks.** De Vries says more than once that he has "no data to back it up." Both men state their biases openly: Medeiros says he'll "die on the hill of WordPress"; de Vries is launching a paid WordPress education product. The critic Medeiros had publicly rebutted is discussed without being named or present. **Bottom line.** A well-reasoned opinion piece with honest caveats, whose most useful claim — that the market-share numbers were never solid — cuts against optimists and pessimists alike. ## 📖 Main Content Medeiros opens on the sentiment he's been criticizing for months: people declaring they're done with WordPress and leaving for AI tools, headless CMSs, and static sites. Is it shiny-object chasing or real pressure from a better product? De Vries starts by shrinking the frame. Everyone in the conversation is in "a very, very tiny bubble," and every metric used to size the larger one — market share, WordPress economy revenue — has always been an assumption. With AI maturing enough to replace some core work, people inclined to chase novelty find it easy to leave. But he thinks they skip a step: learning what WordPress can do, and what the new tool would let them do *within* WordPress. Drawing large conclusions from small, noisy declines is "not being honest with the data." Measured with the same lens, other CMSs aren't surging either. His model: the bottom of the market is peeling off toward AI-generated sites and markdown-file systems like Astro, while WordPress itself matures into something usable in many ways beyond presenting a site. Add a lack of real entrepreneurship among agencies that never had to niche down because work was abundant, and it *looks* causal. He doesn't think it is — the installed base is too big to flip that fast. The struggling agencies and plugin businesses, "and let's not beat around the bush, there's quite a few," mostly never profiled themselves. On why the numbers are so hard to pin down, Medeiros draws a podcasting parallel: RSS's openness is great for distribution and terrible for analytics, which is why big media pushed toward YouTube and Spotify. De Vries agrees and adds a vacation anecdote: without his books, he spent time commenting on Instagram and TikTok and felt the pull of instant feedback. WordPress has no central feedback system. Measurement is further weakened by performance plugins that remove the generator tag, optimizers that obscure plugin and theme paths, and static exports. He mentions Yoast publishing careful caveats about the data. Medeiros recounts his earlier public criticism of a large WordPress content creator for what he considered fear-mongering about WordPress moving too slowly, and restates his bias: WordPress's value is its totality — community, education, meetups, WordCamps, open source — not just the software. De Vries's first question about the critic is "what is he gaining by having this opinion." He expects a transformation, not a collapse, driven by 7.0's native AI interaction and continuing through 7.1–7.3 over roughly a year. WordPress is slower because it serves every scenario from bloggers to enterprises. The community will become "more business-driven and less of an open source," though the base will stay open source, because few people are good at both. They discuss a builder, Mark, who moved to agentic workflows. Medeiros thinks many will come back once the Abilities API matures, since a supported ecosystem beats "you just sitting there with your Claude Code terminal." De Vries suggests Mark was still learning WordPress rather than fully invested, and argues WordPress is already close. He uses Miles, a beta tool by Andy, to build sites and report bugs. Leaving WordPress means losing extensibility and core features — users, plugins, themes, the REST API. Worried about legacy code? Write your own bootstrapper; WordPress doesn't stop you. He sees WordPress as a data warehouse that can use custom tables like WooCommerce does, closer to Laravel's composed approach than people allow. Around 80% of the sites he builds couldn't reasonably be built and maintained by AI because the oversight overhead is too high. He's preferred the command line for ten years, but "it's not for everyone and it's certainly not for my end clients." On the plumber or HVAC business spinning up its own AI site, "danger" is the wrong word. Yes, some revenue will leave. But that market never paid much, the need for multiple editors and stability pulls complexity back in, and projects are growing, not shrinking. De Vries's cousin in that trade can build WordPress sites but declined AI because he wouldn't be able to fix problems and has no guarantee the AI could. The guild segment: announced May 4 2026 at withinwp.com, the guild aims to be the digestion layer for agencies that are working with information more than a year out of date. Learn WordPress, the developer docs, and Make Core hold the information, "but ain't nobody got time for that." He wants to be paid for it and will stay out of agency-growth coaching. The approach is principles first. His *Make WordPress Fast* course is 25 modules of 4–6 lessons of about 4–5 minutes each; he calls it "a boring course" deliberately — understanding what caching is and where it lives, rather than which button to press in Perfmatters, WP Rocket, or NitroPack. He's now building a Cloudflare-for-WordPress course, the hardest to keep current. Medeiros asks how to tell whether a big company's guidance deserves adoption, recalling Google AMP's short life. De Vries's test: is it a set of standards or a proprietary library? AMP was the latter and duplicated what WordPress could already do. Cloudflare's agent-readiness checker collects existing agreed protocols, which he's comfortable with, noting Cloudflare carries over 20% of internet traffic. He recommends APO (Automatic Platform Optimization), a full-page cache at the edge for $5 a month, noting Varnish, Bunny CDN, and Akamai offer similar approaches. The episode closes with pointers to withinwp.com and the WP Minute membership. ### 🗒️ Notes #### ⏱️ [00:00:00](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=0) Framing: leaving WordPress - Medeiros's question: are defectors chasing something new, or reacting to a better product? #### ⏱️ [00:01:04](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=64) The data was never solid - Market share and WordPress-economy figures "have always been an assumption." - Other CMSs aren't gaining at a different rate by the same measures. #### ⏱️ [00:03:12](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=192) The bottom of the market moves - Low-budget users shift to AI-generated sites and markdown systems (Astro named). - "I have no data to back it up… it's my opinion." #### ⏱️ [00:05:18](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=318) Positioning, not software - Agencies have had easy work since about 2010–2011 and never had to niche down. - Struggling businesses "largely" haven't profiled themselves. #### ⏱️ [00:07:24](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=444) Why measurement is hard - RSS/podcasting parallel: open distribution, poor analytics. - Generator-tag removal, path rewriting by optimizers, and static output all hide WordPress. #### ⏱️ [00:11:37](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=697) Fear-mongering and bias - Medeiros's earlier public criticism of a creator; states his own bias openly. - De Vries: "what is he gaining by having this opinion." #### ⏱️ [00:13:44](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=824) Transformation via 7.x - Native AI interaction in 7.0; roughly a year through 7.3. - Community becoming "more business-driven and less of an open source," with an open-source base. #### ⏱️ [00:16:53](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=1013) Agentic builders and the Abilities API - Medeiros: people will return once the Abilities API matures. De Vries: "Not if… this is a when." - Tools named: Miles (beta, used on five sites), WPVibe, WP God Mode. #### ⏱️ [00:20:02](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=1202) WordPress as a data warehouse - About 80% of his sites aren't practical to build and maintain with AI. - Custom tables, own bootstrapper, Laravel comparison. #### ⏱️ [00:22:08](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=1328) The bottom market and upmarket drift - "Danger" is the wrong word; some revenue leaves; complexity returns with a second editor. - Agencies that can't process projects under $50K. #### ⏱️ [00:25:17](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=1517) The Guild - Announced May 4 2026; "It is now August 4th" at recording. - About 30 pieces of content per release; paid; principles first; no agency-growth coaching. #### ⏱️ [00:30:29](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=1829) Principles over buttons - *Make WordPress Fast*: 25 modules; teaches what caching is, not which setting to toggle. - Cloudflare-for-WordPress course in progress. #### ⏱️ [00:34:45](https://www.youtube.com/watch?v=PD5DYYJRmS0&t=2085) Standards versus products - AMP as the cautionary tale; the test is standards versus proprietary library. - Cloudflare's agent-readiness checker collects existing protocols; APO at $5/month. ## 📚 References - Primary: the episode — https://www.youtube.com/watch?v=PD5DYYJRmS0 - withinwp.com (guest's site and the Guild announcement) - The WP Minute — thewpminute.com - Named, not linked: Cloudflare APO; Cloudflare's agent-readiness checker ("is my site agent ready"); Miles; WPVibe; WP God Mode; Perfmatters; WP Rocket; NitroPack; Yoast's writing on market-share measurement; Learn WordPress; developer.wordpress.org; Make WordPress Core. - Not verified here: the "about 44%" market share and the claimed dip, both cited from memory in conversation. Check W3Techs before repeating. ## 🗣️ Paraphrase 1. Lots of people say they're leaving WordPress for AI and newer tools — is that real or just novelty? 2. We live in a bubble, and the market-share numbers everyone cites have always been estimates. 3. The cheapest end of the market is moving to AI site generators; that's real but small in revenue terms. 4. Agencies are struggling because they never had to explain why clients should pick them. 5. Open-web technology is hard to measure, and optimizations increasingly hide that a site runs WordPress. 6. WordPress is changing because of its native AI features, slowly, since it has to support every kind of user. 7. People drawn to chat-based building will probably come back as WordPress's AI tooling matures. 8. Most real sites are too complex to build and maintain with AI alone. 9. Small businesses may build their own AI sites, but that market never spent much anyway, and projects are getting bigger. 10. De Vries is building a paid guild that turns each WordPress release into principles-first lessons. 11. To judge a big company's recommendation, ask whether it promotes standards or its own product. ## 📰 Newsletter **Subject: "Nobody actually knows the numbers"** The most useful line in [this WP Minute episode](https://www.youtube.com/watch?v=PD5DYYJRmS0) comes early. Asked whether WordPress is really losing ground to AI builders, Remkus de Vries says the market-share figures we all argue about "have always been an assumption." That's inconvenient for everyone. It deflates the "WordPress is dying" posts, and it deflates the triumphant rebuttals. Sites that strip the generator tag, rewrite paths through WP Rocket or NitroPack, or ship static output are increasingly invisible to the tools that count them. What de Vries offers instead is a positioning argument. The low end of the market — the plumber who'll let ChatGPT build the site — is leaving, but that segment "was never going to spend much time in the WordPress economy anyway." What's actually hurting agencies is that since around 2010 they never had to explain why a client should hire *them*. Meanwhile, he knows agencies that can't take a project under $50K because discovery alone costs that much. His standards test is worth stealing: when a big company tells you how to do something, is it promoting shared protocols or its own library? AMP failed the test. Cloudflare's agent-readiness checker, he argues, passes. Two things to keep in mind. It was recorded on August 4, so it predates the Automattic board drama. And both men are open about their stakes — Medeiros will "die on the hill of WordPress," de Vries is launching a paid guild. They say so, which is more than most "WordPress is dead" videos do. ## ✅ Action Items **DevRel** - Borrow the "principles first, buttons second" teaching approach for workshops. Explaining where caching lives outlasts any plugin's settings screen. - Use the standards-versus-product test when evaluating vendor AI guidance for content. **OSPO** - Measurement of open-source adoption is weak by design (RSS parallel). Consider what honest usage signals WordPress could publish without adding tracking. **Community** - The claim that the community is becoming "more business-driven and less of an open source" deserves discussion in community teams: how to keep the open-source base healthy as commerce grows. **WordPress subteams** - *Core AI* — de Vries and Medeiros both put the Abilities API at the center of whether agentic builders return. Useful external validation for the roadmap. - *Core Performance* — his principles-first performance teaching and APO recommendation align with core's performance guidance; note that generator-tag removal and path rewriting also hide WordPress usage from market-share counts. - *Training* — the gap de Vries names ("ain't nobody got time" for Make Core posts) is a Learn WordPress opportunity: per-release digests for agencies. - *Marketing* — a narrative for the "WordPress is losing" debate: the numbers are uncertain, the low end is moving, and the value is the whole ecosystem. - *Hosting* — edge caching (APO and equivalents) as a default recommendation. - *Community* — WordCamps and meetups as part of the value proposition Medeiros describes. - *Core*, *Core Program Team*, *Mobile*, *Playground*, *Support*, *Design*, *Documentation*, *Photos*, *TV*, *Polyglots*, *Accessibility*, *Meta*, *CLI*, *Tide*, *Openverse*, *Themes*, *Plugins* — no direct action. **The WPOCC** - The debate about community sentiment and fear-mongering is a governance and communications question the collective could address with better shared data. **FAIR** - Better, verifiable ecosystem metrics (installs, active sites) through federated distribution would address the "we have no clue" measurement problem directly. ### → Things 3 - [ ] Look at withinwp.com's Guild page as a comparison point for Pantheon agency education - [ ] Check Cloudflare's agent-readiness checker against courtneyr.dev ## 🧪 Hypotheses 1. **Detection undercounts WordPress.** A sample of known WordPress sites run through market-share detectors will show a measurable miss rate from generator-tag removal and optimization plugins. Testable against a known list. 2. **Defections cluster at the low end.** Sites leaving WordPress for AI builders will be disproportionately small, single-editor sites. Testable with migration data from hosts. 3. **Differentiated agencies fare better.** Agencies with a defined niche will report more stable revenue in 2026 than generalists. Testable by survey. 4. **Principles-first training retains value longer.** Course content built on concepts will need fewer revisions per year than tool-specific tutorials. Testable by tracking update frequency. 5. **Abilities API maturity predicts agentic adoption within WordPress.** Growth in plugins registering abilities will correlate with more WordPress-based AI building. Testable against directory data. ## 📊 Methodology **Design:** a two-person podcast conversation. Opinion informed by practice, with no data presented. **Evidence:** personal experience (building sites with Miles, agency consulting over ten years, conversations with dozens of agency owners), anecdotes (a cousin in the trades, a builder who moved to agentic tools), and remembered figures (about 44% market share, Cloudflare's 20%+ of traffic, the $50K project threshold, $5/month APO). **Self-assessment:** the guest repeatedly flags that he has no data. That's the most important methodological feature and it's handled honestly. **Limits:** no counter-voice; the critic discussed isn't present; recorded seven weeks before release; both speakers have commercial stakes (the WP Minute membership, the Guild). ## ❓ Questions - What would a trustworthy measure of WordPress usage look like, given how much is hidden? - If the low end leaves and projects move upmarket, what happens to the pipeline of new WordPress developers who used to start on small sites? - Is "more business-driven, less open source" a prediction or a recommendation, and what does it mean for volunteer contribution? - How do agencies differentiate when AI makes build skill cheaper? - Does Cloudflare's agent-readiness guidance hold up as a standard or drift toward its own products over time? - How much of the recording's framing changed after the Automattic board events that followed? ## 🔮 Future Research - An audit of WordPress detection accuracy using a known sample. - Host-level migration data on where departing sites go and what size they are. - A longitudinal study of agency revenue by niche versus generalist positioning. - Measuring whether WordPress's AI tooling (Abilities API, MCP adapter) wins back agentic builders. - Comparing principles-first and tool-first training for retention and outcomes. ## 💡 Implications **Practice.** Agencies should work on positioning — why hire us — rather than on reacting to every AI tool. Learning the principles behind performance and caching keeps skills useful as tools change. **Policy.** Governance discussions should avoid treating uncertain market-share data as proof in either direction. **Security.** Mostly not discussed. The nearest point: de Vries's cousin declined an AI-built site because he couldn't fix it and had no guarantee the AI could. Maintainability is a security property too. **Education.** A strong case for per-release, principles-first education for working professionals — a gap that Learn WordPress and commercial providers could both fill. **Funding.** Education of agencies is being built as a paid product; the community might consider how to make release digests freely available too. ## 🧠 Learning Styles Assessment - **Visual** — weak: a talking-heads podcast. - **Aural** — strong: a conversational audio format that works without video. - **Verbal** — strong: extended explanations and analogies (apples, AMP, RSS). - **Kinesthetic** — moderate: concrete tools to try (APO, the agent-readiness checker, Miles). - **Logical** — moderate: a multi-factor argument, candidly unsupported by data. - **Social** — strong: dialogue between two long-time community members. - **Solitary** — moderate: long-form listening suits individual reflection. - **Naturalistic** — moderate: sorting market segments (bottom, middle, enterprise) and technology types (standard versus library) is a classification exercise. **Reading level:** spoken, about grade 10, with WordPress-insider vocabulary (Abilities API, APO, custom tables). ## 📊 Accessibility Assessment Mapped to WCAG 2.2 AA, evaluated as published on YouTube. 1. **Captions (1.2.2)** — auto-generated only. The guest's name is rendered "Remiss" throughout, several product names are garbled, and a stuck-audio artifact produced a long run of "pl pl pl" at [00:27:20]. *Fail.* 2. **Audio description (1.2.5)** — no essential visual content. *Pass in substance.* 3. **Text alternatives / description** — not assessed; the description wasn't captured. 4. **Keyboard operation (2.1.1)** — inherited from the YouTube player. *Pass.* 5. **Navigation (2.4.x)** — a 40-minute episode with no chapters in the capture. *Fail.* 6. **Readable language (3.1.x)** — conversational with jargon. *Partial.* 7. **Speaker identification** — unlabeled speaker turns (">>") in a two-person conversation. *Fail.* **Top fix:** a corrected transcript with speaker labels and correct names, plus chapter markers. ## 🌍 DEIB Assessment **Cultural breadth.** Transatlantic (US host, Dutch guest). Examples are Western small-business markets. **Inclusive language.** Generally respectful; the discussion of the unnamed critic stays on the argument rather than the person, though "what is he gaining" questions motive. **Neurodiversity.** Long, digressive answers ("Sorry for the monologue") make it hard to follow without chapters. **Representation.** Two men discussing the community; no women's voices, and the people discussed (Mark, James, Andy) are men. **Psychological safety.** Friendly and good-humored; both speakers disclose bias, which models openness. ## ✅ Logical Validity ### Strong reasoning - Questioning the data before interpreting it is the right first move. - Separating correlation from causation: several trends coincide, and he explicitly says they only look causal. - The standards-versus-library test is clear and generalizable. - Repeated, explicit disclosure of lacking data. ### Fallacies 1. **Motive over argument.** *Where:* [00:12:39] "what is he gaining by having this opinion." *Why it fails:* it addresses the critic's incentives rather than his claims. *Fix:* engage the claim that WordPress is moving too slowly directly. 2. **No-true-Scotsman risk.** *Where:* [00:16:53] the suggestion that Mark "never was fully in WordPress." *Why it fails:* dismissing a defector as never fully committed makes the "people who really know WordPress stay" claim unfalsifiable. *Fix:* take the defector's stated reasons at face value. 3. **Anecdote as trend.** *Where:* the $50K agencies, the cousin, the 80% figure. *Why it matters:* useful color, but none of it is systematic. *Fix:* frame as experience. 4. **Inconsistent use of data.** *Where:* he calls market-share data unreliable, then says other CMSs aren't gaining by the same measure. *Why it matters:* if the data is unreliable, it's unreliable for both claims. *Fix:* acknowledge the symmetry. ### Tone Warm, candid, occasionally self-deprecating. Pro-WordPress, openly so. ## 🛡️ DARVO Analysis **Does the source perform DARVO?** No full pattern. Medeiros's account of criticizing a creator for "fear-mongering" and de Vries's question about the critic's motives lean toward the **Attack** move (questioning motives rather than engaging claims), but there's no denial of wrongdoing and no victim reversal. **Is DARVO described?** No. **Assessment:** low. The motive-questioning is worth noting as an argumentative habit, not a DARVO sequence. ## 🎯 Conclusion **Bottom line.** A thoughtful, candid conversation that reframes "is WordPress losing to AI?" as three separate claims: the numbers were never reliable, the low end is moving, and agencies' real problem is positioning. De Vries's principles-first teaching philosophy and his standards-versus-product test are the most reusable ideas. The weaknesses are the usual ones for opinion podcasts — no data, anecdotes, and motive-questioning of an absent critic — and it was recorded before the Automattic governance events that followed. **Next steps.** 1. Use the standards-versus-product test when evaluating AI-web guidance. 2. Treat market-share arguments with the skepticism de Vries recommends. 3. Watch whether the Abilities API draws agentic builders back, as both speakers predict. ## 🧱 Refactor Appendix ### Literature Split Pad **L1 — The data was never solid.** Market-share figures are assumptions; detection is increasingly defeated. Source: WP Minute, Remkus de Vries, recorded 2026-08-04. **L2 — The bottom is moving, projects are moving up.** Low-budget sites leave for AI; complexity returns with a second editor; agencies struggle to take work under $50K. **L3 — Positioning is the crisis.** Agencies never had to differentiate since about 2010. **L4 — Standards versus product.** AMP was a library; Cloudflare's checker collects existing protocols. **L5 — Principles first.** Teach what caching is, not which setting to toggle. #### Atomic Split Pad **A1 — Unreliable numbers can't settle a trend in either direction.** If you distrust the data for the decline, distrust it for the rebuttal. Links: [[Measurement honesty]]. **A2 — When a big company prescribes, ask: standard or product?** Links: [[AMP lessons]], [[Vendor guidance]]. **A3 — Teaching principles outlasts teaching buttons.** Links: [[Principles-first pedagogy]]. **A4 — Build skill stops differentiating when tools make it cheap.** Positioning becomes the product. Links: [[Agency positioning]]. **A5 — The second editor brings back complexity.** Single-user AI sites break down once collaboration is needed. Links: [[Multi-user CMS value]]. #### Fleeting Queue - Confirm spellings for Mark, James, and Andy/Miles from the show notes. - Check the "44%" market share and dip against W3Techs. - Note the recording date (August 4) wherever this is cited. <!-- deep-read:sections end --> 📄 **Full text:** [[Full Document Contents/Articles/Is WordPress Really Losing Ground to AI.md|Is WordPress Really Losing Ground to AI? by WP Minute]] ## Highlights - actual state and for all the metrics that we have used to gauge how large the larger bubble is meaning whether we're looking at WordPress's market share or how much money is going through the WordPress economy all those types of metrics and ideas and thoughts about that we have to start with understanding that those have always been an assumption. We have no clue. So what I see happening is we are entering the phase where AI is sort of I wouldn't ([View Highlight](https://read.readwise.io/read/01m31zckxkft1n8dppkyjddvn3)) - Note: Atom A1 — unreliable numbers can't settle a trend in either direction. Market-share and WordPress-economy figures "have always been an assumption." Cuts against both the "WordPress is dying" and the rebuttal camps. - Tags: wordpress-market-share, atom/atomic/h1 - context because I remember the days of Google AMP and everyone's like we got to do Google AMP and like Google AMP like now we got to do this and that was like whatever the hell that lasted for two two not even two years I don't even remember you know and I just feel like big company comes out they go this is the way and then we all go okay let's go this way and then we scale it back. How do you make draw those lines in the sand if that makes sense? >> It it does make sense and it it's a very valid question. Uh look at it from a perspective of are they promoting a set of standards or are they promoting a set ([View Highlight](https://read.readwise.io/read/01m31zcs0wmkp30kqe2pa26670)) - Note: Atom A2 — when a big company prescribes, ask: standard or product? AMP was a library; Cloudflare's agent-readiness checker collects existing protocols. - Tags: web-performance, atom/atomic/h1