Pages cited by AI systems more often put a direct answer near the top, according to an analysis of 193,507 AI responses. It is a useful clue for publishers, not a recipe for winning citations: the data show an association, not that changing a page will make AI cite it.
Watch Desk analysis
What happened
Search Engine Journal describes an analysis by TJ Digital’s TJ Robertson, drawing on 4,167 questions and eight weekly snapshots from 27 August to 21 October 2026. In one comparison, cited pages began with a direct answer more often than pages retrieved for the same AI answer but not cited: 23% versus 14%.
In a separate blind-rating round, cited pages were more likely than Google top-ten pages that AI had not retrieved to open with a direct answer: 20% versus 5%. That comparison covered 195 pairs of pages. The article says the first comparison’s page count and the AI model used for ratings had not been identified as of 28 October.
Key findings
- An answer near the top was more common on cited pages
The pattern appeared in both comparisons, though the results do not establish that it caused citations. - The comparisons are not interchangeable
The 23% versus 14% result compares pages found by the same AI answer; the 20% versus 5% result compares cited pages with Google top-ten pages the AI did not retrieve. - The data have a defined, commercial sample
The analysis uses client questions tracked by an SEO agency, covering 78 companies and nine business models. It is not a representative census of the web.
Why it matters
For a publisher or business, the practical takeaway is modest but actionable: check whether a page answers the question it is meant to address near the beginning. That is a test worth trying, not a guaranteed route into an AI answer.
The distinction matters because the study compares existing pages rather than testing edits against a control group. It cannot show that moving an answer higher will increase citations. The sample also comes from one agency’s client data, and the limited blind ratings leave important methodological details open.
Our read
This is a useful pattern, not an SEO spellbook. The finding gives publishers something concrete to examine, while the study’s own limits are a reminder not to confuse “more common among cited pages” with “the thing that made them get cited”.
What to watch
- Whether a later study tests page changes against a control group.
- Whether the analysis publishes the missing details on its first comparison and rating model.
- Whether the pattern holds across other datasets, question types and AI systems.
Discussion spark: Should publishers move direct answers to the top of a page on the strength of this association, or wait for experiments showing that the change actually earns more citations?
Sources and evidence
- Source update (28 October 2026, 20:42 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.