Guide
Google and Microsoft give opposite AI search advice
By Oğuzhan Tüsen @oguzhanglotierPublished
Two companies whose search engines both power AI answers say different things about schema, about structure, and about writing for machines. Here are all six sentences, and the more useful finding underneath them.
The three topics, in their own words
Every quote below was read from the vendor’s own page on 2026-07-31. Links are at the bottom so you can check them rather than take our word for it.
Structured data and schema
“Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add.”
Microsoft
“Schema is a type of code that helps search engines and AI systems understand your content.”
Breaking content into pieces
“There's no requirement to break your content into tiny pieces for AI to better understand it.”
Microsoft
“Bulleted lists, numbered steps, and comparison tables break complex details into clean, reusable segments.”
Writing for machines
“You don't need to write in a specific way just for generative AI search.”
Microsoft
“Success starts with content that is fresh, authoritative, structured, and semantically clear.”
Read them precisely, because the gap is smaller than it looks
It is tempting to call this a flat contradiction and move on. It is not quite one. Google says schema is not required. Microsoft says it helps. A thing can help without being required, so both sentences can be true at the same time, and each vendor is describing systems it actually runs.
The structure row narrows further once you read the whole Google page rather than the sentence that suits the argument. Google objects to breaking content into tiny pieces for AI, and in the same guide says: “People generally appreciate it when web pages are organized by paragraphs and sections, along with headings that provide a clear structure.” That is not far from Microsoft at all. Both want a page a reader can follow. The disagreement is narrower than the framing suggests and it is specifically about schema.
What neither of them wrote is the sentence the industry sells: add FAQ schema and you will get cited by AI. That instruction is stronger than Microsoft’s and directly against Google’s, and it is the version that ends up in audit reports as a red item.
So the finding is not that a vendor is lying. It is that the advice being sold is stronger than the evidence behind it, and that no claim about what “AI” rewards survives without naming the engine.
Why they differ at all
Different retrieval stacks. Google’s AI Overviews are generated on top of the existing Search index, reached by Googlebot, which is why Google keeps pointing back at ordinary content quality. Microsoft grounds Copilot in Bing and separately sells retrieval tooling where document structure genuinely changes results, so structure is a natural thing for it to recommend.
Perplexity is a third shape again: it owns a purpose-built answering index and runs its own search on every question. Three architectures, one set of “best practices” sold across all of them.
This is not permission to ignore Microsoft
Microsoft’s recommendations are additive rather than prohibitions. Lists, tables and one-to-two sentence answers are cheap to write, Microsoft says they help on its surface, and Google nowhere says they hurt. A page can follow both vendors without conflict.
There is also independent evidence on one piece of it. The peer-reviewed GEO benchmark (KDD 2024, 10,000 queries) measured adding quotations, statistics and cited sources each lifting visibility 30 to 40%, and keyword stuffing decreasingit 30.3% for the top-ranked page. That supports making a claim easy to lift and attribute, which is nearer Microsoft’s framing than to nothing. It says nothing at all about schema, which is the specific thing Google denies.
What we changed because of this
Our own audit now attributes every recommendation to the vendor that made it. It reads “Microsoft recommends this, Google says it is not required” rather than picking a side and presenting it as consensus. That is more words and a less confident-sounding report, and it is the only version we can defend when a customer opens the other vendor’s documentation.
It is also the clearest argument we know against a single AI visibility score. A number that blends engines which disagree about the mechanism cannot tell you what to do next, however precise it looks.
Questions
Does schema markup help AI search?
Microsoft says it helps its systems understand your content. Google says it is not required for generative AI search and that there is no special schema.org markup you need to add. Both are describing their own systems and both can be true, because helping and being required are different claims. What nobody at either company wrote is that schema gets you cited.
Should I add FAQ schema for AI visibility?
That specific instruction comes from neither vendor. Schema still earns rich results in classic Google Search, which is a real and separate reason to add it. If the goal is being named in an AI answer, Google's own guide puts unique and useful content above every structural suggestion it makes.
Why do Google and Microsoft give different advice?
They run different retrieval stacks. Google's AI Overviews sit on top of the Search index. Microsoft grounds Copilot in Bing and also sells retrieval tooling where structure genuinely matters. Neither is wrong about its own system, and the error is averaging them into one set of rules that matches neither.
So should I ignore Microsoft's advice?
No. Its recommendations are additive rather than prohibitions, so a page can follow both vendors without conflict. Lists, tables and concise answers are cheap and Microsoft says they help on its surface. The thing you cannot do is state that they are what AI rewards, without saying which AI.
Is there any independent evidence either way?
On extractability, yes. The peer-reviewed GEO benchmark (KDD 2024, 10,000 queries) measured adding quotations, statistics and cited sources each lifting visibility 30 to 40%, and keyword stuffing decreasing it 30.3%. That supports making a claim easy to lift and attribute, which is closer to Microsoft's framing. It says nothing about schema.
Sources
Google, AI optimization guide. Microsoft, optimizing your content for inclusion in AI search answers. The GEO benchmark, arXiv 2311.09735, KDD 2024. All read on 2026-07-31.
Related: who actually blocks AI crawlers, whether llms.txt does anything, and what an AI visibility score actually measures.
Do this automatically
Get Glotier Solo
Everything on this page is one measurement, taken by hand, on one day. Solo runs it for your product every day and writes the work it points to.
- The same buyer questions re-asked every day, so you see the moment an answer changes
- Every source page behind each answer, which is where the work actually is
- An article written from what it measured that day, ready for you to publish
- The subreddits and threads worth answering, with a reply drafted for each
Track 3 products and up to 150 buyer questions across ChatGPT, Gemini and Perplexity. $39/month.
Cancel any time. Not for you? Email us within 7 days of a charge and we refund it in full. Refund policy
Not ready to pay? The check is free, with no card and no account. Run it on your own product.