The essentials in 60 seconds
Structured data (Schema.org) is not mandatory for appearing in Google's generative AI: its official May 2026 guide confirms this. It remains useful for traditional rich results (reviews, FAQs, products, events) and for removing ambiguity about your entities. In other words: a way to speed eligibility, not a magic ranking driver.
We still see client teams piling up JSON-LD in the hope of a ranking jump. That is not how it works. This article separates what triggers a rich result, what helps AI, and what serves no purpose, using Google's primary sources and measured 2026 figures.
Is structured data required for Google's AI?
No. Google's AI guide (15 May 2026) is explicit: structured data is not required for generative search. Its thesis fits into one sentence — optimizing for AI means doing SEO. Search Engine Journal summarizes it: AEO and GEO are still SEO.
Google does not require structured data for AI answers. Schema remains useful for eligibility for conventional rich results; it must match the visible content. Google announced the launch of AI Overviews and AI Mode in France on July 22, 2026. AI Overviews appears when Google considers a generated answer useful for the query; it does not replace all conventional results.
llms.txt, chunking, writing “for AI”: myths to put to rest
Google Search does not use llms.txt: neither a benefit nor a penalty. The May 2026 guide says so in its mythbusting section. No content chunking requirement, no need to write specifically for AI, and structured data is not required for answer generation. These “tricks” waste time.
| Common belief | What Google says (2026) |
|---|---|
| llms.txt improves Google rankings | False — Google Search does not read it (neither bonus nor penalty) |
| Content must be “chunked” for AI | False — no chunking requirement |
| Write “special AI” pages | Unnecessary — good SEO is enough |
| Schema is required for AI | False — useful but not required for generation |
An honest caveat: llms.txt may serve other crawlers or engines, but there is no evidence of an effect on Google. If you publish it, do so for those third-party use cases, not in expectation of a Google gain. To put these components in context, see our AEO / SEO / GEO comparison.
Why AI traffic is fragmenting (and what that changes)
Because every AI search engine cites a different web. ChatGPT citations overlap with only around 12 % of Google SERPs (Higoodie, 2026), and ~80 % of LLM citations do not rank in Google's top 100 (Ahrefs, 15 000 queries). Your AI visibility therefore does not mechanically follow your organic rankings.
- overlap between ChatGPT citations and Google SERPs
- ~12 %
- Higoodie (2026)
- LLM citations absent from Google's top 100
- ~80 %
- Ahrefs (2026)
- AI Search traffic conversion rate (vs. ~2,8 % organic)
- ~14,2 %
- Seer Interactive (2026)
AI Overviews increasingly cite pages outside the top 10: traditional ranking alone is no longer enough.
- % of citations from the top 10
Source: Ahrefs / AI Overviews (2026)
View chart data
| Reference | % of citations from the top 10 |
|---|---|
| Mid-2025 | 76 |
| Early 2026 | 38 |
What we measure on our accounts: only ~11 % of domains are cited by both ChatGPT and Perplexity. The operational consequence is clear — work on citability (factual clarity, explicit entities) alongside rankings. Details in our GEO/AEO essentials.
Falling CTR and distorted measurement: what to watch
Search Console provides a generative AI performance report for AI Overviews and AI Mode impressions. Google reports a global rollout as of August 31, 2026, with access limitations depending on the property and impression volume. This data is also included in the Web report: do not add it to that report's total. Compare identical periods, countries, devices and aggregations. The chart is aggregated by property, or by URL when a URL filter is applied; the page table may therefore have a different total. Google's counting rules do not establish automatic double-counting of AI and conventional impressions. A CTR decline calls for an analysis of traffic and queries, not a blanket adjustment.
- GSC: AIO + organic impressions added together, no filter to isolate them.
- GA4: check the AI Assistant channel and session source/medium data; a lost source cannot be inferred from the Direct channel.
- ChatGPT accounts for ~87 % of AI referral traffic (Conductor, 2026).
- Perplexity is the 3rd source (~7,7 %, Statcounter, May 2026).
What to do with structured data in practice in 2026
Deploy Schema where it enables a real rich result, rather than everywhere. Prioritize types eligible for enhanced display, validate them, then measure the impact on rich result impressions rather than rankings. The rest — llms.txt, “AI” pages — contributes nothing for Google.
- Map pages with transactional or rich-content intent (products, FAQs, reviews, events, articles).
- Implement the corresponding JSON-LD and validate it (Rich Results Test).
- Make entities explicit (Organization, Person, sameAs) to remove ambiguity.
- Strengthen E-E-A-T: firsthand experience, identified author, sources — see Google's guide.
- Track rich result eligibility, not “AI ranking”: there is no dedicated Schema driver.
On quality, the May 2026 core update ended on 2 June, followed by a global spam update on 24 June. Danny Sullivan reiterated it: Google judges quality, not the production method. Low-value mass content receives the lowest rating — structured data will not save it. Our services and our AI Overviews France report detail implementation.
Frequently asked questions
Does structured data improve my Google rankings?
Not directly. Structured data makes a page eligible for rich results (reviews, FAQs, products), but is not a ranking factor in itself. Google's AI guide (May 2026) confirms it is not required for generative search either. It clarifies your entities, nothing more.
Should you publish an llms.txt file for Google?
No. Google Search does not use llms.txt: neither a bonus nor a penalty, according to the May 2026 AI guide's mythbusting section. The file may serve some third-party crawlers or engines, but there is no evidence of an effect on Google. Do not rely on it to gain Google visibility.
Is Schema necessary to appear in AI Overviews?
No. Structured data is not required for generating AI answers. It remains useful for conventional rich results and for disambiguating your entities. AI Overviews and AI Mode have been available in France since July 2026: work on content quality and clarity.
Why does my ChatGPT traffic not follow my Google rankings?
Because AI search engines cite a different web. According to Higoodie (2026), ChatGPT citations overlap with only around 12 % of Google SERPs, while Ahrefs finds that ~80 % of LLM citations do not rank in the top 100. Work on AI visibility alongside organic rankings.
Does Google penalize AI-written content?
Not on principle. Google judges quality, not the production method (a position reaffirmed by Danny Sullivan). But low-value mass content receives the lowest rating in the Quality Rater Guidelines, and the June 2026 spam update targets these abuses. Added value remains the criterion.
How can I measure the real impact of my structured data?
Track rich-result eligibility and the associated impressions in the search appearance report. For AI impressions, check the generative AI performance report: they are also included in the Web report, and the totals must not be added together. In GA4, check source/medium data and the AI Assistant channel; Direct is not a measure of AI traffic.
