Neither is required for Google’s AI features (Google Search Central, Dec 10, 2025). Ahrefs’ study of 1,885 pages that added schema found no clear citation uplift on pages that were already cited (Ahrefs, May 11, 2026). Its llms.txt data counts requests to the files, not whether they change what AI mentions: 97% of the files in its 137,000-site study got no requests in May 2026 (Ahrefs, June 15, 2026).
If you already have them, keep them, but don’t expect them to get you named. This post shows the numbers, what Google says, what happened on our own site, and what we would work on instead.
What are schema and llms.txt, in plain terms?
Schema is structured data, usually written as JSON-LD, that describes a page to machines: this is a business, this is its name, this is its address. An llms.txt file is a proposed plain-text file that is meant to point AI tools to the most important content on a site.
Schema has been around for years and is part of normal search engine work. llms.txt is newer. According to Ahrefs, “no major AI platform has ever committed to reading it” (Ahrefs, June 15, 2026).
Did adding schema get pages cited more by AI?
No clear uplift in the study of already-cited pages. Ahrefs tracked 1,885 pages that added JSON-LD between August 2025 and March 2026 and compared them with 4,000 control pages, measuring 30 days before and after (Ahrefs, May 11, 2026). Against the controls, citations changed by:
- Google AI Mode: +2.4%
- ChatGPT: +2.2%
- Google AI Overviews: −4.6%
Ahrefs found the first two statistically indistinguishable from zero. It reports the AI Overviews figure as statistically significant. Its summary: “Adding schema produced no major uplift in citations on any platform.”
That significant AI Overviews gap doesn’t show that schema caused it. In AI Overviews, the treated and control pages were both already on a steep downward trajectory before schema was added, and Ahrefs says it can’t tell from this data alone whether schema or something else explains the gap. Pages that add schema often change other things at the same time, so the study can’t fully separate the two.
There is one more caveat, and Ahrefs states it. Every page in the test already had 100 or more AI Overview citations before the change. So the test says nothing about a page that AI hasn’t cited yet. It also measures citations, which is a different outcome from a business being recommended or named, so it doesn’t show that schema is useless for every site.
Do AI assistants read llms.txt?
Very few requests suggest they do. Ahrefs looked at 137,210 domains and found that 38,360 of them, 28%, had a valid llms.txt file. Of those files, 97% received zero requests in May 2026. Of the requests that did happen, AI retrieval bots, the ones behind AI search, made 1.1% (Ahrefs, June 15, 2026).
Two caveats, both Ahrefs’ own. First, the domains come from Ahrefs Web Analytics customers, who skew more technical than the web at large, so 28% is probably an upper bound. Second, the study measured requests to the files in one month, not whether an AI tool uses what it fetches. So we would not say llms.txt files are “ignored by AI”. The accurate version is that 97% of them got no requests in May 2026.
Do you need special SEO for Google’s AI Overviews?
No. Google’s page on AI features says: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” It also says: “You don’t need to create new machine readable files, AI text files, or markup to appear in these features. There’s also no special schema.org structured data that you need to add.” (Google Search Central, last updated Dec 10, 2025)
What a page does need, per Google, is to be “indexed and eligible to be shown in Google Search with a snippet”. Its best practices include allowing crawling, keeping important content in text, and making sure any structured data matches the visible text on the page. In other words, ordinary SEO is the starting point.
This is the strongest source in this post, because it comes from the company that runs those features. It covers Google’s AI features only. It says nothing about ChatGPT or Perplexity.
What happened on our own site?
Two separate observations, and neither is a test. answersky.com has one Organization schema block and no llms.txt file: our /llms.txt returned a 404 when we checked on Oct 4, 2026. Separately, our first scan on Sep 30, 2026 gave a score of 0/100 on ChatGPT, Gemini and Perplexity, and Answersky was named in none of the answers. We ask the models behind ChatGPT, Gemini and Perplexity, via their APIs with web search, so the apps themselves can answer differently. The full record is in our public log.
The two checks are on different dates and weren’t a before-and-after or a controlled comparison. They tell us nothing about what either file does, and we don’t draw any wider conclusion from them.
Should I remove my schema or llms.txt?
No, removing them isn’t the point. Keep accurate schema that matches what visitors can see, which Google lists as a best practice. If you have an llms.txt, it is a small file to leave in place. Just don’t treat either as a fix or expect it to get you named.
The −4.6% AI Overviews result was statistically significant, but the study can’t establish that schema caused it (see above), so it isn’t a reason to remove schema either. It is one test, on pages that were already heavily cited.
What should I work on instead?
Three things, each with a source. None is a promise that you will be named.
- Let AI search crawlers in. OpenAI says sites that opt out of OAI-SearchBot “will not be shown in ChatGPT search answers, though can still appear as navigational links” (OpenAI, checked Oct 4, 2026). Open your robots.txt and check. Our post on why ChatGPT doesn’t recommend a business shows how.
- Write plain text that answers the buyer’s question. Put the question in a heading and the answer in the first line under it. Google’s best practices list making sure “important content is textual”.
- Get mentioned on the sites AI already cites for your buyers’ questions. Ahrefs found that in Google’s AI Overviews, branded web mentions had a 0.664 correlation with brand visibility across 75,000 brands, against 0.218 for backlinks (Ahrefs, May 26, 2025). That is a correlation, and it is for Google, so it doesn’t prove that mentions cause ChatGPT to name a business.
The third one is what we are testing on ourselves. Our results, good or bad, go in the public log. You can also read how we take each measurement on how it works.
How do I know if a change worked?
Change one thing at a time, and trust a result only when 2 scans agree. A single before-and-after comparison can easily be noise. We ask each question 3 ways, in 3 different wordings, once each, because that gives a steadier number than one answer (why we ask each question 3 ways). And we count a change only when 2 scans agree (how the score is built).
A change you make on a Monday and judge on a Tuesday from one answer tells you very little. Two scans that agree tell you a lot more.
Rather than add another file, see what ChatGPT and Perplexity say about you now.
Free Quick Scan: 6 of your buyers’ questions on ChatGPT and Perplexity, 12 answers, and who AI named instead of you. Report within 3 business days. See pricing.