Will removing schema markup hurt my AI visibility?

Removing schema markup is unlikely to change how AI engines describe or cite you. Keeping it is unlikely to help. Both halves of that answer come from the same evidence.

Google's guide to its generative AI features says you don't need new markup or machine-readable files to appear in Google Search, including its AI capabilities. Its AI features page adds that there are no special optimizations needed for AI Overviews or AI Mode. Eligibility comes from being indexed and eligible to show with a snippet.

The best experiment available points the same way. Ahrefs tracked pages that added schema against matched controls and found citations barely moved. You can read the full breakdown in our digest of the Ahrefs schema study.

So schema sits outside the levers that matter for AI answers. That cuts both ways. If adding it does nothing measurable, removing it is not a fix either. Google says existing markup does no harm. The honest position: leave working markup alone, and stop treating it as an AI project.

Two sources carry most of the weight here.

Google's own statements

• Structured data "isn't required for generative AI search", per Google.
• No AI-specific schema.org type exists for Google's AI features.
• FAQ rich results were deprecated in May 2026, and Google removed the FAQPage documentation in June 2026. See FAQPage schema and its current status.

Google's statements are assertions, not measurements. They are still the best available evidence for Google surfaces.

The Ahrefs matched-control test

Ahrefs studied 1,885 pages that added schema, each compared with three matched controls. It found "no major uplift in citations on any platform." The results by surface:

None of this supports schema as a citation lever. The AI Overviews drop also doesn't prove schema hurts, which is why ripping it out isn't the answer either. For the wider picture, see does schema markup help AI citations.

There is one honest boundary on the Ahrefs result. Every page in its treated set already had more than 100 AI Overview citations before schema was added. Ahrefs studied pages that AI was already citing heavily.

That leaves a question open. Does schema help a page that AI never cites get into the answer at all? Nobody has published a test of that. It is untested, not refuted.

Say you run a small accounting firm that no AI engine names today. The Ahrefs study tells you little about your exact situation. It also gives you no reason to expect schema to change it. When evidence is missing, the sound move is to avoid spending effort in either direction and focus on the levers with documented mechanisms.

Some cases justify touching your schema. None of them are about AI citations.

• The markup states wrong facts. If your JSON-LD lists an old price or a discontinued product, fix it. Contradictions between your schema, pricing page and footer are a plausible source of wrong AI answers. Check with the free JSON-LD checker.
• FAQPage markup you added for rich results. Those rich results are gone. Leaving the markup causes no harm, so removal is a housekeeping choice, not a visibility decision. FAQ schema after the deprecation covers the details.
• Markup still powers classic rich results. Some schema types still drive rich results in regular Google Search. Deleting those can cost you features in classic results, separate from AI.
• Maintenance burden. If a plugin generates schema nobody owns, stale markup is a liability. Either assign an owner or remove it.

The same logic applies to old pages with outdated facts. See whether to delete or noindex old pages.

Put your effort where the evidence is strong. As of October 2026, the levers that hold up are access levers. They work by removing blocks.

1. Check per-bot robots.txt rules. Make sure OAI-SearchBot, PerplexityBot and Claude-SearchBot are allowed. GPTBot governs training, not ChatGPT search.
2. Check the firewall and CDN. A firewall can block a crawler that robots.txt allows.
3. Put key facts in the raw HTML. Vercel found ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. Facts that appear only after JavaScript runs may be invisible to those crawlers.
4. State specific facts in plain text. A price, a number or a one-line definition on a fetchable page gives an engine a sentence it can quote. Compare the approaches in structured data vs plain-text facts.

To see where you stand today, the free checker on this site shows whether AI engines name you for three buyer questions. It is a snapshot, not continuous tracking.

Is FAQ schema deprecated? Yes, for Google rich results. Google deprecated FAQ rich results in May 2026, removed the FAQPage documentation in June 2026, and dropped Search Console API support in August 2026. This applies to the markup, not to question-and-answer content on your page. Writing clear questions and answers is still fine for readers. The markup just no longer earns a rich result, and no evidence shows it earns AI citations.

Should I add schema to new pages for AI search? Not for AI reasons. Google says structured data isn't required for its generative AI features, and Ahrefs found no major citation uplift after adding it. If a schema type still powers a rich result you want in classic Google Search, that's a valid reason. For AI answers, spend the effort on crawler access and on stating key facts in the page's raw HTML.

Does schema help with ChatGPT or Perplexity? No engine documents using schema as a citation signal. In Ahrefs' matched test, ChatGPT citations rose 2.2% after schema was added, which Ahrefs did not treat as a major uplift. Perplexity has published nothing on it. For ChatGPT search, the documented controls are OAI-SearchBot permission in robots.txt and letting OpenAI's published crawler IPs through your firewall or CDN.

Can wrong schema cause AI engines to state wrong facts about me? Nobody has measured that directly. What is plausible is that contradictory facts across your own site, including a schema block, give engines conflicting versions to pick from. AI platforms do get specific facts wrong about specific brands. Fixing contradictions in your markup, pricing page and about page removes one possible source of errors, and it's fully within your control.