FAQ schema died on May 7th, 2026. Most SEO guides never noticed.

---

QUICK ANSWER

Google removed FAQ rich results from Search on 7 May 2026, and plenty of guides published after that date still tell you to add FAQ schema. That is one of six pieces of standard SEO advice checked here against Google's own statements, court testimony, and the actual studies people cite. A table of contents, topic-cluster architecture, word count targets, readability scores and bounce rate all turn out to be weaker than advertised, or measuring something that was never a ranking factor. What survives afterward is short.

---

Most SEO advice circulating right now includes at least one thing Google has already killed, said explicitly does not matter, or that traces back to no real study at all. Nobody updates the guide once it is published. Here is what actually checked out, one tactic at a time.

FAQ schema stopped working on 7 May 2026

Google's own developer documentation carries a deprecation notice, added 8 May 2025, stating plainly that FAQ rich results would stop appearing in Search "starting May 7, 2026." They did. Google's own changelog logged the removal the next day.

The markup itself is not broken. Google says it may still use FAQPage structured data to understand a page. What it no longer does is show your questions and answers directly in the search result, which was the entire reason anyone added it in the first place. A guide written before May 2026 that still recommends building an FAQ block for the rich result is describing a feature that no longer exists.

The removal was staged rather than sudden, which is worth knowing if you are trying to date when a piece of advice went stale. The notice went up a full year before anything changed. Search Console's own reporting and its Rich Results Test dropped support for the feature the following month. API support followed two months after that. Anyone who checked Google's own documentation at any point in that year would have seen the date coming.

A table of contents was never quite what it looked like

The story usually told is that Google requires a table of contents to generate the "jump to" links some search results show. That gets it backwards. Google's own 2009 guidance did once recommend building a page with anchors and a linking table of contents to enable exactly that feature. Since then, passage ranking has let Google surface a specific answering section using its own understanding of the page, without needing a manually built map of it.

What never checked out is the specific numbers used to sell the practice today. The "18% CTR uplift" and "cuts bounce rate by 40%" figures that circulate alongside table-of-contents plugins trace to no named study, no sample size, and no link to primary research anywhere this research could find. They appear on plugin marketing pages and nowhere upstream of them.

A well-structured page with clear headings still helps a reader and gives Google something to work with. A table of contents block bolted on to chase a specific, uncited percentage is decoration, and decoration is not free: it is one more thing rendered above the actual answer a visitor came for, on the same page this whole series has argued should load fast on a slow connection.

Topic clusters come with a different invented statistic every time you look

Search for the traffic increase hub-and-spoke content architecture supposedly produces and the number changes depending on which vendor is selling it that week: 30%, 40%, 55%, 119%, 254%, 1000%, and a separate HubSpot case study claiming a single blog went from 500 to 190,000 monthly visitors. None of these share a methodology. None cite an independent source. That is not six data points converging on one finding. It is six marketers each picking their own best case.

The one piece of this that does trace to a real, dated Google statement points the other way. Asked directly about varying internal anchor text between related pages, John Mueller said, in a Webmaster Central hangout on 24 June 2020, that it is "not something where you would see a visible effect in search." Linking your own pages to each other still matters for the ordinary reason that it helps Google find and understand them. Carefully varying the exact wording each time, as a tactic in itself, does not.

That distinction gets lost because it is a genuinely small one to state precisely, and a much bigger one to sell. "Link your pages sensibly" does not fill a content-strategy retainer. "Build a documented hub-and-spoke architecture with tracked anchor-text variation" does, and the invoice arrives whether or not the underlying claim survives contact with Google's own engineers.

Word count and readability scores measure nothing Google says it uses

Google has said directly that word count "is not a quality factor, not a ranking factor." The study most often cited against that, Backlinko's analysis of 912 million blog posts, found longer posts accumulate more backlinks and social shares. It never established that length itself moves a page up the results, and the distinction was pointed out directly at the time: correlating with links is not the same claim as correlating with rank.

Readability scores fare no better. A study of 756,297 pieces of content across 30,000 search queries found no relationship between Flesch Reading Ease and where a page ranked. The average reading level of a top-30 result sat at an eleventh-grade level regardless of whether it ranked first or thirtieth.

Bounce rate is not the signal, and a new site has no signal yet either way

Google has been unusually direct about this one. Gary Illyes said outright in 2015 that bounce rate is not used in ranking. John Mueller repeated it in a June 2020 hangout, and Illyes separately dismissed dwell time and similar signals in a Reddit AMA as "generally made up crap."

What Google does use is aggregated click data, through a system called NavBoost, confirmed under oath by Google's own VP of Search during the 2023 US antitrust trial: roughly a hundred billion clicks a year, built up over about thirteen months. The same testimony confirmed NavBoost cannot rank a page that has no click history yet, which is every new site, for a while, no matter how good the content is.

That is a genuinely different problem from bounce rate, and worth not confusing with it. Bounce rate was never a ranking input, at any age of site. Click history is a real input, and a new site simply has not accumulated any yet. No amount of on-page optimisation changes that; only time and enough real visitors do.

What actually still holds up

Structure your headings so a reader, and Google, can follow the page without needing a manually built map. Link your own pages to each other in plain language, because that helps people and machines find them, not because varying the words each time earns anything extra. Write for a person, since no readability score is buying you anything with a search engine that has already checked.

And keep going. A new site's actual problem was never search volume in the first year: it is stopping before month four, which pays the entire cost of publishing and collects none of what compounds after.

---

Find out what you were told to do

Elevates is a design and engineering studio.

Send us your current site, or the last content brief another agency gave you, and we will tell you plainly which of the six things above they are still selling you.

Both answers are free, and one of them might save a line item nobody is checking.