There is a popular piece of advice going round: add more schema markup and you will start appearing in Google's AI Overviews. Google has published the opposite in its own documentation. This page shows you exactly what Google says, what structured data is genuinely worth doing, and the single rule that trips most sites up — your markup has to match what a human can actually see on the page.
Source: Google Search Central, "AI features and your website", checked October 2026.
Where to start
- Someone sold you an "AI schema" package and you want to know if it was real? Read Step 1, then ask them for the citation.
- You already have schema and just want to know if it is set up correctly? Go to Step 2 and run the match check.
- You have no schema at all and want to know whether to bother? Step 3 tells you the three cases where it genuinely pays.
What Google actually wrote
Google's documentation page "AI features and your website" covers how content shows up in AI Overviews and AI Mode. Three lines from it do most of the work here.
On special markup, Google writes: "There's also no special schema.org structured data that you need to add."
On new files, which is the same advice people ignore when they rush to add an AI-specific text file: "You don't need to create new machine readable files, AI text files, or markup to" appear in these features.
And the broad one: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary."
That is about as direct as Google's documentation gets. There is no AI schema type. There is no markup that flips a switch. Anyone selling you one is selling you something Google has said in writing does not exist.
Be careful about what this does not say, though. It does not say schema is useless. It does not say schema is ignored. It says there is nothing extra you need for the AI surfaces specifically, on top of normal good practice. Those are different claims, and the difference is the whole point of this page.
The rule that does matter: make it match
Google's same page lists, as a best practice, "Making sure your structured data matches the visible text on the page." This is the one that breaks real sites.
It breaks in boring ways. A plugin writes a Product price of 29.99 into the markup while the page shows 34.99 because someone changed it in the shop and not in the SEO field. A recipe says it serves four in the markup and six on the page. A review block claims an aggregate rating that appears nowhere a reader can see. Each of these is a mismatch, and a mismatch is worse than having no markup at all, because it is a machine-readable statement that contradicts your own page.
Mismatch checker
Paste one value from your structured data and the matching value a visitor sees on the page. This runs entirely in your browser; nothing is sent anywhere.
Run that for every value you have marked up: price, availability, rating, author, publish date, serving size, duration. The ones that drift are almost always the ones a human edits often and a plugin caches quietly.
When schema is genuinely worth your afternoon
Structured data earns its keep in three situations, none of which are "AI Overviews".
One: rich results in classic search. This is the real payoff and it always has been. Review stars, recipe cards, FAQ expanders, product price and stock, event dates, breadcrumbs. These change how your listing looks on an ordinary results page, which changes how many people click it. That is a measurable gain you can see in Search Console.
Two: removing ambiguity about facts. If your page is about a person, a product or an organisation whose name is shared with something else, marking up the entity plainly helps any machine reading the page work out which one you mean. This is ordinary good hygiene, not a trick.
Three: feeding your own systems. Plenty of teams use their own structured data for internal search, for a product feed, or to keep a site consistent. That is a perfectly good reason on its own.
Should I add schema here?
Answer two questions about the page you have in mind.
- Stop paying for anything described as "AI schema" or "AEO markup". Google says in writing there is no such requirement.
- Pick your five most important pages and compare every marked-up value against what a visitor sees.
- Fix mismatches at the source. If the markup is typed separately from the page content, it will drift again.
- Keep the schema types that earn rich results in classic search. Drop the ones that earn nothing.
- Spend the time you just saved on the thing Google does ask for: content a person actually finds useful, in text, easy to reach.
Two honest caveats, because this topic attracts overclaiming on both sides.
First, "Google says you do not need it" is not the same as "Google ignores it". Structured data helps any system parse a page, and Google has never said otherwise. The claim this page is pushing back on is the specific one — that there is a special schema that gets you into AI Overviews. There is not.
Second, this covers Google's AI surfaces, because Google is the only major engine that has published guidance this plain. ChatGPT, Perplexity and the rest do not document how they weigh structured data, and nobody outside those companies can tell you. If someone quotes you a precise figure for how much schema lifts your citations in ChatGPT, they are guessing.
Sources: Google Search Central, "AI features and your website" (checked October 2026). Quotations are taken verbatim from that page. Nothing in this guide is sent to a server — every tool runs in your browser.