Guide

Does AI cite AI-written content?

Most new web pages now have some AI writing in them, and most business owners wonder whether that helps or hurts. Here is what the largest studies measured about AI-written pages in Google's rankings and AI Overviews, what Google's rules actually say, and where AI writing does cost you.

Quick answer

Yes, and how much AI wrote a page barely matters. Ahrefs found 87.8% of pages cited in Google AI Overviews mixed human and AI writing, and the correlation between AI share and citation order was 0.017, effectively zero. Google says it rewards quality "rather than how content is produced", but mass-produced pages that add nothing break its spam rules.

What are the key facts about AI-written content and AI citations?

FactDetailSource
Cited pages are mostly mixedOf pages cited in AI Overviews, 3.6% were "pure AI", 8.6% "pure human" and 87.8% a mix (top three citations from 1 million AI Overviews)Ahrefs (third-party)
AI share doesn't predict orderCorrelation between a page's AI-content share and its citation order in AI Overviews: 0.017Ahrefs (third-party)
Fewer pure-human pages citedNew web pages overall: 2.5% pure AI, 25.8% pure human, 71.7% mixed (900,000 pages)Ahrefs (third-party)
Rankings don't care eitherAcross the top 20 results for 100,000 keywords (600,000 pages), correlation between AI share and rank was 0.011; 86.5% of top-ranking pages had some AI contentAhrefs (third-party)
Fully AI pages can rank first5.3% of pages ranking 1–3 were 100% AI-generated; 82.2% of top-3 pages were under 50% AI (331,000 pages)Ahrefs (third-party)
But heavy AI costs reachIndexation fell from 49.28% for low-AI pages to 40.35% for very-high-AI pages; low and moderate AI pages got 2–3× the impressionsAhrefs (third-party)
Google's positionIts focus is "on the quality of content, rather than how content is produced" (February 2023)Google
Where Google draws the lineUsing generative AI to make many pages without adding value for users may violate its scaled content abuse policyGoogle
Fact-check requiredGoogle says to manually fact-check and review all AI-generated content, including titles, meta descriptions, structured data and alt textGoogle
No fake authorsMade-up names, AI headshots or false credentials for authors are deception and a signal of a low-quality pageGoogle
AI features want what AI can't writeGoogle asks for non-commodity content, not content that "could easily be produced by a generative AI model"Google

Do AI Overviews cite AI-generated pages?

Yes, constantly, and almost always pages that mix human and AI writing. Ahrefs, an SEO software company, took 1 million searches that showed an AI Overview, pulled the top three cited links from each, and ran the roughly 500,000 pages it had on record through its own AI-content detector. Ahrefs

  • 3.6% of cited pages were classified as pure AI.
  • 8.6% were pure human.
  • 87.8% were a mix of the two.

Of the mixed pages, 11.2% showed minimal AI use (1–10% of the text), 44% moderate use (11–40%), 24.7% substantial use (41–70%) and 7.9% dominant use (71–99%).

The correlation between how much of a page was AI-written and where it was cited in the AI Overview was 0.017: no relationship at all.

For comparison, Ahrefs ran the same detector on 900,000 newly published pages across the web and found 2.5% pure AI, 25.8% pure human and 71.7% mixed. Ahrefs So pure-human pages make up about a quarter of new pages but under a tenth of AI Overview citations. Ahrefs titled its study "AI Overviews Cite AI-Generated Content More Than Human Writing".

Read this carefully. AI detectors are estimates, not proof, and these are one company's detector and one company's sample. The safe conclusion is the narrow one: Google's AI does not avoid AI-assisted pages. It does not show that adding AI writing earns citations.

Does AI-generated content hurt Google rankings?

Not by itself, in two large Ahrefs studies. Because Google's AI features are built on its core Search ranking systems, this matters for AI citations too. Google

600,000 pages. Ahrefs took 100,000 random keywords and the top 20 results for each. 4.6% of ranking pages were pure AI, 13.5% pure human and 81.9% mixed, and 86.5% of top-ranking pages had some AI content. The correlation between AI share and ranking position was 0.011. The very top positions had slightly less AI content, which Ahrefs read as at most a minor preference. Ahrefs

331,000 pages. A second study looked for where AI writing does cost something. Ahrefs

  • 5.3% of pages ranking in positions 1–3 were 100% AI-generated, so fully AI pages can rank at the top.
  • But 82.2% of top-3 pages were under 50% AI.
  • Average AI share rose only slightly from position 1 (27.1%) to position 10 (30.9%).
  • Indexation dropped: 49.28% of low-AI pages were indexed against 40.35% of very-high-AI pages.
  • Reach dropped: low and moderate AI pages received two to three times the impressions of high and very-high AI pages.

Ahrefs' summary: Google is not against AI content; it is against bad content, and the two overlap a lot of the time.

What does Google say about AI-written content?

That it judges the result, not the tool, and that mass-producing pages to rank is spam whatever tool made them.

Quality, not method. In February 2023 Google wrote that its ranking systems aim to reward original, high-quality content showing expertise, experience, authoritativeness and trustworthiness, and that its focus "on the quality of content, rather than how content is produced" has helped it deliver reliable results for years. Appropriate use of AI is not against its guidelines; using it to generate content primarily to manipulate rankings is. Google

Scaled content abuse. Google's spam policies define scaled content abuse as generating many pages primarily to manipulate rankings rather than help users, "no matter how it's created". The first example listed is using generative AI tools to generate many pages without adding value for users. Others include scraping or rewording other sites and stitching content together without adding value. Google

Check everything. Google's guidance on generative AI content says models predict likely words rather than retrieve facts, so outputs can contain errors, and it is critical to manually fact-check and review all AI-generated content before publishing. That includes titles, meta descriptions, structured data and image alt text, because those appear in results. Google

Don't invent people. The same page says fabricating creator profiles, with AI-generated headshots, made-up names or false credentials, is deception, and any deception makes a page untrustworthy. It suggests disclosing how AI was used where a reader would reasonably ask "how was this created?"

Say something AI can't. Google's guide to its AI features asks for non-commodity content based on in-depth experience, and warns against recycling what others have said or content that "could easily be produced by a generative AI model." Google

When does AI-written content cost you citations?

When it says nothing new, says something wrong, or is published in bulk. This is our reading of the data and Google's rules together.

  • When it is generic. An AI engine has no reason to cite a page that says what the model already knows. Google says as much: content that a generative model could easily produce is the commodity content its AI features pass over. The GEO study found the edits that raised a page's share of AI answers were added statistics, quotations and cited sources, the things a model cannot invent honestly. See What the Princeton GEO study found.
  • When it is wrong. A made-up price, licence rule or statistic on your page can be repeated in an AI answer about you. Google says to fact-check all AI output; we say treat every number as unverified until you have checked it against your own records.
  • When it is mass-produced. Fifty near-identical city pages or service pages written by AI to catch searches are the textbook case of scaled content abuse. In Ahrefs' data, very-high-AI pages were also indexed less often, and an unindexed page cannot be cited.
  • When the author is fake. A made-up expert byline is, in Google's words, deception. Use real names of real people who checked the page.

How should a business use AI to write pages that get cited?

This is our advice, built on the Ahrefs data and Google's guidance. Use AI for the drafting; supply everything that makes the page worth citing yourself.

  1. Start from your own facts. Before you prompt anything, write down your prices, timelines, service area, licence numbers, guarantees and one or two real results. These are what no model can make up correctly.
  2. Let AI draft the structure. Question headings, a first-sentence answer under each, tables for prices and options. AI is good at this and Google does not object to it.
  3. Replace every generic sentence with a specific one. "We offer competitive pricing" becomes "A standard 2,000 sq ft re-roof in Tampa runs $14,000–$19,000 (our 2026 jobs)."
  4. Fact-check every claim. Google says to manually review all AI output, including titles, meta descriptions, schema and alt text. Remove anything you cannot back up.
  5. Add a real person. A named owner or technician who reviewed the page, with their real credentials. Never an invented author or AI headshot.
  6. Add a quote or a source. A line from a customer who agreed to be quoted, or a link to the code, manufacturer page or study behind a claim.
  7. Publish fewer, better pages. One strong page per real customer question beats fifty thin ones, and keeps you clear of the scaled content abuse policy.
  8. Disclose when it helps. If a reader would reasonably ask how a page was made, say so in a line: "Drafted with AI, checked by [name]."

What should be on my AI-written content checklist?

CheckDone whenSource
Own facts firstEvery page carries at least three facts from your own recordsOur advice
Fact-checkedEvery number, rule and name was checked by a person before publishingGoogle
Metadata checkedTitle, meta description, structured data and alt text were reviewed tooGoogle
Real authorThe byline is a real person with real credentialsGoogle
Not commodityThe page says something a model could not produce on its ownGoogle
No bulk pagesNo sets of near-identical pages made to catch searchesGoogle
IndexedEach page shows as indexed in Search ConsoleAhrefs (third-party)
Quote or sourceEach page has a real quote or a linked source for its key claimGEO paper
Disclosure where expectedPages a reader might ask about say how AI was usedGoogle

How do I know if my AI-assisted pages are working?

Check two things for each page: whether Google has indexed it, and whether any AI engine cites it for the question it answers. A page that is not indexed after a few weeks is a warning sign; Ahrefs found heavily AI-written pages were indexed less often.

The Prompt Simulator runs one question across ChatGPT, Gemini, Perplexity and Google's AI answers and shows which sources each one quotes. Our free AI visibility check shows where your business is named today. For what to fix on each page, see The 12-point GEO audit checklist.

Sources

  1. AI Overviews Cite AI-Generated Content More Than Human Writing — Ahrefs. Read Oct 5, 2026.
  2. 74% of New Webpages Include AI Content (Study of 900k Pages) — Ahrefs. Read Oct 5, 2026.
  3. AI-Generated Content Does Not Hurt Your Google Rankings (600,000 Pages Analyzed) — Ahrefs. Read Oct 5, 2026.
  4. Google Doesn't Punish AI Content; It Punishes Bad Content (331k Pages Studied) — Ahrefs. Read Oct 5, 2026.
  5. Google Search's guidance about AI-generated content — Google Search Central. Read Oct 5, 2026.
  6. Google Search's Guidance on Generative AI Content on Your Website — Google Search Central. Read Oct 5, 2026.
  7. Spam policies for Google web search — Google Search Central. Read Oct 5, 2026.
  8. Google's Guide to Optimizing for Generative AI Features on Google Search — Google Search Central. Read Oct 5, 2026.
FAQ

Common questions

Does Google penalize AI-generated content?

Not for being AI-generated. Google says its focus is on the quality of content rather than how it is produced, and Ahrefs measured a 0.011 correlation between AI share and ranking across 600,000 pages. Google does penalize generating many pages with AI to manipulate rankings, under its scaled content abuse policy. Google Ahrefs

Do AI Overviews prefer AI-written pages?

Ahrefs' data suggests they do not avoid them. Pure-human pages were 25.8% of 900,000 new pages but 8.6% of pages cited in AI Overviews, and 87.8% of cited pages mixed human and AI writing. The correlation between AI share and citation order was 0.017, so AI writing did not predict being cited higher. Ahrefs

Can a 100% AI-written page rank at the top of Google?

Yes. In Ahrefs' study of 331,000 pages, 5.3% of pages in positions 1–3 were fully AI-generated. But 82.2% of top-3 pages were under 50% AI, and very-high-AI pages were indexed less often (40.35% vs 49.28%). Ahrefs

Do I need to disclose that I used AI?

Google does not require a label on every page. Its guidance suggests disclosing AI use where a reader would reasonably wonder how the content was created, and says fabricating authors, such as AI headshots or made-up credentials, is deception. Google

Will ChatGPT or Perplexity ignore AI-written pages?

Neither company has said it treats AI-written pages differently, and there is no public study of either engine's handling of AI content that we could verify. What both reward, in the studies we have read, is specific, sourced material a model can quote. Ahrefs

Is it OK to use AI to write location pages for every city I serve?

Only if each page has real, different information for that city. Google's scaled content abuse policy names using generative AI to create many pages without adding value as an example of spam. Fifty pages that differ only by city name fit that description. Google

See whether AI answers quote your pages.

Enter your domain. The free AI visibility check shows where ChatGPT, Gemini, Perplexity and Google's AI answers mention you, and which of your pages they draw on.