Product Features Overrode a #1 Listicle Spot in Claude's Top Recommendation

Conventional GEO wisdom assigns the #1 slot in a listicle a 50% citation probability and drops the #2 slot to 30%.

Jabez Reuben reports placing a client in the #1 spot of a third-party listicle in May, where product features overrode the #1 listicle spot in Claude's top recommendation.

Published on an external domain rather than the client's site, the article listed the client first and a competitor lower down the page.

When Claude generated answers citing that exact third-party URL, the lower-ranked competitor captured the #1 recommendation.

Claude parsed the query context to identify which specific attribute satisfied the prompt: the listicle credited the competitor with Feature X and the client with Feature Y, making Feature X the winning match regardless of page order.

Because industry data only reflects campaigns built around top-slot placements, a planned follow-up test places a target brand in the last spot of a listicle to isolate attribute matching from rank order.

Furthermore, deleting the competitor from the published third-party article failed to remove the brand from Claude's answers, which kept recommending the competitor via the same citation URL.

The persistence exposed citation age — an LLM cache delay where models retain an older snapshot of a page and take time to refresh their understanding after live edits (sometimes this takes months).

Because cached retrieval lags behind on-page edits, Reuben avoids chasing brand insertions or edits inside existing third-party articles, judging the ROI too low.