# Funnel Position Predicts Which Content AI Search Extracts Not all content is equally exposed to AI search. The exposure is predictable, and the predictor is funnel position. **Top-of-funnel content is the most extractable, because it answers a question completely and briefly** — which is precisely the format an overview can restate, and precisely the case where the reader has no remaining reason to click. Middle- and bottom-of-funnel content survives because it requires context the summary doesn't carry: your specifics, your product, the decision the reader is actually making. SignLab is the clean experiment. They ran a programmatic SEO play — one page per sign, per language, each with a short video. A parent searching for how to sign "stop" got a complete answer and left. When Google began surfacing the clips in AI Overviews, **rankings and impressions held steady while clicks went to near zero**. The content ranked exactly as well as before; it just stopped being a reason to visit. The important part is the failed remedy. Their first instinct was to make the programmatic pages *better* — more detail, more depth. It did not work: "AI Overviews just takes your content and don't really give a lot back." Enriching extractable content produces better extraction, not better retention. What recovered the business was building deliberate middle- and bottom-of-funnel coverage — content whose value cannot be separated from the site it lives on. So the diagnostic is not "is my content good" or "does it rank." It is: **can a reader who gets a faithful summary of this page still need the page?** If no, the page is a donation to the aggregator, and optimizing it harder increases the size of the donation. <!-- atomicity: 1 idea ✓ --> ## Why This Matters For **Courtney's own publishing**, this sorts the portfolio directly. "What is a block theme?" is extractable and always will be. "Here is what broke when I migrated this specific site to a block theme, and what I'd do differently" is not — the specificity is the moat. The same split applies to WordPress tutorial content generally: reference material is exposed, lived experience and opinionated judgment are not. For **DevRel work**, it argues against the reflex to publish comprehensive introductory documentation as an audience-acquisition strategy. That material is now the most efficiently harvested category on the web. It may still be worth writing for its own sake — but it should be budgeted as a public good, not as a channel. For **honest measurement**, it is the strongest case the vault has that rank and impressions can be perfectly healthy while the business outcome goes to zero. The metrics didn't lie; they measured something that stopped mattering. ## Counterpoints to Hold - **One company, unusual shape.** Short video assets answering atomic questions are close to the maximally extractable case. A site whose top-of-funnel content is long, argumentative, or brand-inflected may be far less exposed. - **Funnel position isn't the only predictor.** Freshness, licensing posture, brand strength, and whether the answer is a *fact* versus a *judgment* all matter. Funnel position is a good first cut, not a complete model. - **Retreating from top-of-funnel has a cost.** It's the content that reaches people who don't know you yet. Abandoning it to protect clicks may protect the metric and shrink the audience — SignLab kept theirs running. - **The floor may keep rising.** Nothing guarantees middle- and bottom-of-funnel content stays un-extractable as models get better at synthesizing across sources and personalizing. This is a description of the current frontier, not a stable safe harbor. ## Links - [[3. Resources/Permanent Notes/Republishing Into a Rival Index Recovers Traffic an Aggregator Intercepted]] — the sibling remedy from the same source: this note says which content is exposed, that one says what to do with it - [[3. Resources/Permanent Notes/Search Visibility and Business Value Need Separate Metrics]] — evidence-for: the sharpest case available, where every visibility metric held steady and every value metric collapsed - [[3. Resources/Permanent Notes/AI Search Inherits the Trust and Fragility of the Sites It Cites]] — the other end: the overview is defunding the sites it depends on for the answers it gives - [[3. Resources/Permanent Notes/A Metric Is Vanity Only Against Your Own Outcome Data]] — applies-in: impressions weren't vanity until they decoupled from clicks; the decoupling is the whole event - [[3. Resources/Permanent Notes/Education Is the Only Marketing That Works With Developers]] — contrasts: educational content is exactly what's most extractable, so the strategy still holds but the channel economics behind it don't - [[3. Resources/Permanent Notes/Perm · The ecosystem will bifurcate between 'human-authored' and AI-scale producers]] — extends: extraction pressure is one of the forces doing the splitting ## Open Questions - Does the pattern hold for text-heavy tutorial content, or is video uniquely extractable because the overview can embed the asset itself? - Is there a measurable middle case — content that's extractable but where attribution still drives meaningful referral? ## Source - [[3. Resources/Literature Notes/Lit · Losing half of Google traffic to AI Overviews]] — SignLab founder interview, 2026-07-24 - See also: [[3. Resources/MOCs/DevRel MOC]]