On May 15, 2026, Google Search Central published a guide called "Optimizing your website for generative AI features on Google Search" and announced it the same day on the Search Central Blog. It applies whether the site runs on WordPress, Shopify, Wix, Squarespace, Webflow or hand-built templates.
The first thing to do is keep your foundations in place: useful original pages that Google can crawl and index, and structured data that describes what a visitor can actually see on the page. Nothing in the guide asks you to add a new kind of markup.
Google's guide says structured data is not required for generative AI search and that no special schema.org markup is needed. In the same guide, Google says structured data remains useful for applicable Google rich results. Both statements are dated May 15, 2026, and both apply to Google Search.
What Google published on May 15, 2026
Google's announcement describes the new resource as a place to learn about useful, unique content, local, shopping, image and video content, common AEO and GEO misconceptions, early guidance for AI agents, and why SEO best practices remain the base. The audience it names includes site owners, SEOs, developers, publishers, local businesses and online stores that want visibility in AI Overviews, AI Mode and Google Search in general.
The same day, the Google Search documentation updates page recorded the change in one line: "Adding a new guide on optimizing for generative AI features." The full text lives in Google's own page, Optimizing your website for generative AI features on Google Search.
Google also keeps a related page, AI features and your website. It says a page that appears as a supporting link in AI Overviews or AI Mode needs to be indexed and eligible to appear in Google Search with a snippet, and that there are no additional technical requirements or special optimizations for these features.
Google's two statements about structured data
Google's guide puts it this way: "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." It then qualifies the point. Structured data remains useful as part of an overall SEO strategy, because it can support eligibility for Google rich results.
Read the two statements together. The first is about AI Overviews and AI Mode, and it says Google does not ask for an AI-only layer of markup. The second is about the features that still exist in Google Search, such as article and breadcrumb displays, where accurate markup can make a page eligible. Google's general structured data guidelines add that markup should represent visible content and that markup does not guarantee a rich result.
That describes what Google says it needs for its generative AI features. Bing, Apple, Yandex and other AI systems document their own behavior, and none of this is a reason to remove accurate markup that already describes your pages.
What Google says about AEO and GEO
If you have come across AEO and GEO in agency decks and search forums, the terms have simple expansions. AEO is answer engine optimization, the habit of shaping pages so answer-oriented systems can use their information. GEO is generative engine optimization, aimed at visibility, mentions or citations in generated answers. The label GEO was formalized in a 2023 research paper, GEO: Generative Engine Optimization, later published at KDD '24. AEO grew out of SEO and answer-engine marketing communities and has no single agreed definition.
Google's guide defines both terms and then says that, from Google Search's perspective, optimizing for generative AI search is still SEO. In other words, Google treats AEO and GEO as labels for work on visibility in AI search, not as a separate ranking system. Other providers may describe their own systems differently, so the guide settles the question only for Google.
The guide also explains two mechanics. Retrieval-augmented generation, or RAG, means Search systems retrieve relevant pages to support an AI response. Query fan-out means related searches are generated to cover the subtopics of a question. Both rest on Google's existing ranking and quality systems, according to the guide.
What Google says site owners should do
Google's list is short and familiar. Create useful, original, non-commodity content with a distinctive point of view. Organize pages clearly for readers. Use relevant, high-quality images and video. Avoid mass-producing pages for every possible query variation. Keep pages crawlable and indexable, reduce duplicate content, keep local and ecommerce details accurate, and keep applying Search Essentials and SEO best practices.
For a small business, the accurate-details item is the one that touches markup most directly. If your opening hours, address, prices or product availability appear in structured data, they need to match the page and match what the business really offers. A mismatch is a fault whether or not an AI feature is involved.
Google also lists what Google Search does not need. According to the May 15 guide, you do not need llms.txt or similar special AI files, special AI markup or Markdown, content broken into arbitrarily small chunks, AI-specific rewriting, an exact match for every long-tail query variation, inauthentic mentions, or special schema.org markup for generative AI search. Each item is Google's statement about Google Search, dated May 15, 2026.
Make the markup readable for Bing and AI search
Bing's structured data documentation lists Schema.org, JSON-LD, Microdata, RDFa, Microformats and Open Graph, so the same JSON-LD serves both engines. To see Copilot citations, use AI Performance in Bing Webmaster Tools, covered in Bing AI Performance and Copilot citations. Yandex does not support every Schema.org type, so check its semantic markup questions and validator.
To be eligible for ChatGPT search, allow OAI-SearchBot in robots.txt; placement is not guaranteed (OpenAI's help article). For Perplexity, allow PerplexityBot there too (crawler documentation). Read whether AI systems use schema markup for the full comparison.
What to keep doing on your site
Start with the page as a visitor sees it, then check that your markup says the same thing. Open the rendered page source and look for the types your theme or SEO plugin already outputs before adding anything. On WordPress, Shopify, Wix, Squarespace and Webflow, an existing block may already describe the article, the organization or the product, and a second block describing the same item can contradict the first.

Then work through a short routine. Mark up content that is genuinely present and visible. Use Schema.org types and properties that describe the page accurately. Keep titles, dates, authors, images, breadcrumbs, products and business details consistent between the markup and the page text. Use JSON-LD when it is the clearest format for your site, validate the result against Schema.org and the documentation of each provider you care about, and keep sitemaps, internal links and metadata in good order.
Write direct, useful answers for people rather than text built to trigger an AI system. If you decide to allow or block a crawler such as OAI-SearchBot, PerplexityBot or Applebot, treat that as an access decision made in robots.txt, separate from your markup.
The guide Schema for AEO and GEO walks through what to mark up for answer engines. To build the article block shown below, the Article schema generator produces valid JSON-LD from the details you enter.
Before and after markup
Before, a site might add a block built only to influence AI answers, such as an FAQPage entry that restates a line of copy. Google's May 2026 documentation update said the FAQ rich result would no longer appear in Google Search from May 7, 2026, so FAQPage is not added to chase that display. FAQPage remains a valid Schema.org type; the point is that Google offers no feature for it to unlock.
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "Does AI search require special schema?",
"acceptedAnswer": {
"@type": "Answer",
"text": "No special AI schema is required for Google Search."
}
}
]
}After, keep ordinary, accurate article and breadcrumb markup in one graph. The example below uses fictional example.com values. Replace the @id and url values with your page's address, headline and description with the page's own text, datePublished and dateModified with real dates, and the author and publisher with the real people and organization named on the page.

{
"@context": "https://schema.org",
"@graph": [
{
"@type": "BlogPosting",
"@id": "https://example.com/blog/generative-ai-guide#article",
"url": "https://example.com/blog/generative-ai-guide",
"headline": "A Fictional Guide to Generative AI Search",
"description": "A fictional article explaining documented search guidance.",
"image": ["https://example.com/images/generative-ai-guide.jpg"],
"datePublished": "2026-05-15",
"dateModified": "2026-05-15",
"author": {
"@type": "Person",
"name": "Alex Example",
"url": "https://example.com/authors/alex-example"
},
"publisher": {
"@type": "Organization",
"name": "Example Publishing",
"url": "https://example.com",
"logo": {
"@type": "ImageObject",
"url": "https://example.com/images/example-publishing-logo.png"
}
},
"mainEntityOfPage": {
"@type": "WebPage",
"@id": "https://example.com/blog/generative-ai-guide"
}
},
{
"@type": "BreadcrumbList",
"@id": "https://example.com/blog/generative-ai-guide#breadcrumb",
"itemListElement": [
{
"@type": "ListItem",
"position": 1,
"name": "Blog",
"item": "https://example.com/blog"
},
{
"@type": "ListItem",
"position": 2,
"name": "A Fictional Guide to Generative AI Search"
}
]
}
]
}The property rules come from separate places. Google's Article structured data page lists no required properties for article rich results and names headline, image, datePublished, dateModified and author as recommended when applicable. Google's Breadcrumb structured data page requires itemListElement, and each ListItem needs position and name; item is required except on the final crumb. Check the property list for your page type before you publish, because Google's pages can change.
Keep Schema.org vocabulary and Google's feature documentation apart. BlogPosting is a subtype of Article in Schema.org, and mainEntityOfPage is valid vocabulary even though Google's article feature does not require it. A valid type is not automatically a Google feature, and do not paste this example into a page unchanged.
What no markup can promise
Nobody outside a provider can verify every retrieval, ranking, selection or citation decision made inside a closed system. Markup can clarify content and support eligible search features, but it cannot guarantee rankings, AI citations, traffic or inclusion. Valid JSON-LD means the syntax parses and the vocabulary fits; whether a page is indexed and shown is a separate step.
For the same reason, one AI answer is a weak test. A fair comparison changes only the markup, keeps the visible content identical, records dates and prompts, repeats the queries and keeps ordinary rankings apart from AI citations. Keep accurate markup in place for the systems that document a use for it.
Questions people ask
Does Google require special schema for AI Overviews?
Google says no. Its May 15, 2026 guide states that structured data isn't required for generative AI search and that no special schema.org markup needs to be added. Pages still need to be indexed and eligible to appear in Google Search with a snippet.
Should I remove structured data because it is not required?
No. In the same guide, Google says structured data remains useful as part of an overall SEO strategy because it can support eligibility for Google rich results. Bing, Apple and Yandex also document uses of markup for their own search features.
Does Google use llms.txt?
Google's May 15, 2026 guide lists llms.txt and similar special AI files among the things Google Search does not need for generative AI search. That is a statement about Google Search only, so a file you keep for another system is a separate choice.
Are AEO and GEO different from SEO?
Google defines AEO as answer engine optimization and GEO as generative engine optimization. From Google Search's perspective, optimizing for generative AI search is still SEO. Other providers document their own systems, and the labels are industry shorthand rather than standards.
Can JSON-LD guarantee an AI citation?
No provider promises that. Bing says markup alone does not guarantee a rich result, and OpenAI says placement in ChatGPT search is not guaranteed. Accurate markup describes your content; it is not a route to a particular citation.
