Study plan, registered on 2 October 2026. This study has not been run yet. I’m publishing the method before any results exist so it can’t be adjusted to fit them. Results will be added to this page, with the date they were collected.
When an AI engine answers a UK news question, the answer can include a link to a news source, name a source without linking to it, or do neither. Only the first can send a reader to the publisher. This study will count how often each happens.
What will be measured
Each answer will be scored as a whole, with one row per query and engine. An answer with at least one clickable link to a news source counts as linked. One that names a news brand without any link counts as a mention only. One that names no news source counts as neither. It uses the same 30 queries, engines and date as the AI answers study plan, where the full query list is published.
This is a narrower cut of the same run. It’s published as its own study because it answers a different question from the per-brand count.
Limits
Scoring the whole answer doesn’t show which brand got the link. The per-brand results will show that. AI engines also change how they cite sources without notice, so the results will describe one date.
What would show the result was wrong
Running the same queries on the same day and getting a clearly different split between the three outcomes for any engine.
Data
This study will share its CSV with the AI answers study, in a separate column for the whole-answer score.