SEO / Research · Review your setup
Ahrefs’ ChatGPT study points back to original sources. Make your own facts easy to find.
Source published 2026-10-12 · Brief published 2026-10-13 · Updated 2026-10-13
What changed
Ahrefs published an analysis of 66.6 million free, logged-out GPT-5.6 Luna responses and 15 million search queries from August and September. It reported that 14.4% of responses triggered web search. This is a vendor dataset covering a particular model and setup, not a universal description of ChatGPT.
What this actually means
Query fan-out means the additional searches an AI system makes to investigate a question. Those searches, the sources it cites and visits to your website are different observations. My recommendation is to keep them separate in reporting, so a visibility score does not become a claim about commercial impact.
Who should care
B2B teams planning AI-search content around assumptions that every question triggers web retrieval or that a frequently searched domain will necessarily be cited. Teams need evidence from their own buyer questions.
The implications for your team
For your content budget, the useful question is which information your business can explain reliably at its source. Product requirements, supported integrations, implementation scope and honest evidence can help buyers whether they arrive through conventional search, an AI answer or a salesperson. That is a stronger starting point than publishing interchangeable summaries.
Editorial recommendation
Your next useful move
Review one commercial page for missing buyer facts. Add verified answers where they belong and connect them to the relevant evidence. Do not create a page for every imagined AI search variation.
What could change in practice
The immediate change should be more complete and consistent source material. Search or citation changes may follow, may take time or may not occur. Treat them as observations to record, while separately checking whether the revised page helps people evaluate the offer.
How to put it into practice
These are recommended next steps. Confirm the feature or workflow is available in your account before changing production settings.
Collect real evaluation questions
Use sales conversations and support questions to choose a small set about one offer. Include suitability, prerequisites and implementation concerns; document the origin of each question.
- Who owns it
- SEO lead with sales
- Check before moving on
- The set reflects real buyer needs rather than invented search demand.
Audit the authoritative page
Check whether the page answers those questions accurately. Assign missing facts to product or delivery owners. Resolve contradictions between the website and sales documents before adding markup.
- Who owns it
- Content and product owners
- Check before moving on
- Every important answer has an accountable source.
Publish readable evidence
Write direct answers with supporting detail and relevant internal links. Check the rendered mobile page and actual crawl access; mark only structured data that matches visible content.
- Who owns it
- Editor and website owner
- Check before moving on
- People and crawlers can reach the same clear, accurate information.
Observe without overclaiming
Record test prompts, model, date and cited URLs using permitted tools. Track referrals and qualified enquiries separately. Repeat comparable checks before attributing a change to the revision.
- Who owns it
- Analytics owner
- Check before moving on
- The report distinguishes answer presence, citations, visits and sales outcomes.
How to check whether it worked
The page answers verified buyer questions and the team can explain what each visibility metric measures. A citation is useful evidence of discovery, not proof of additional customers. Keep commercial evaluation grounded in observed demand.
What to keep in mind
The study’s model, logged-out setup, dates and prompt mix affect the findings; paid or reasoning modes can behave differently. Search-query and citation analyses use different datasets. Its percentages should not be applied as forecasts for a B2B category.
Read the original
Ahrefs Blog · 2026-10-12
How ChatGPT Does Query Fan-Out (15 Million Queries Studied) ↗