In a Peec AI experiment, the AI systems tracked by the team retrieved YouTube videos with revised descriptions much more often. Daily retrievals rose by 359% for the updated videos and by 17% for those left unchanged. For businesses with an existing video library, that raises a question worth testing: did the new description help, and did more people actually see a link?
What changed in the experiment?
According to Search Engine Land, the team analysed about 60 videos. On 19 October 2026, they rewrote the descriptions of 15 videos and added manual timestamps. The videos themselves stayed the same. The first comparison window ran from 6 to 21 October and the second from 22 October to 4 November.
The changes on 19 October fell within the first window. That window therefore includes days on both sides of the intervention, so it cannot serve as a clean baseline.
What does the 359% increase measure?
Peec AI researcher David Konitzny reported a 359% increase in daily retrievals for updated videos, 17% for unchanged videos and 25% for YouTube overall in the tracking data. The updated group also grew faster than the overall YouTube figures in Google AI Overviews and AI Mode.
The metric counts how often the tracked systems retrieved a video as a potential source. The published figures do not show how often a link appeared in a final answer, how often people clicked it, or how many became customers. The percentages alone also do not show the absolute baseline.
Why the cause remains uncertain
The team changed the descriptions and timestamps together. The experiment cannot separate their effects. Retrievals also rose for YouTube overall, so wider trends could account for some of the change. The public findings do not establish whether another channel or another language would see similar results.
How to test this on your own channel
Start with a few older videos that still answer relevant questions but have thin descriptions. Revise only the description text for one group and leave several comparable videos unchanged. Test timestamps separately. Record the original descriptions, video URLs and change dates, and use comparable periods to review the results.
Choose questions your videos actually answer. If a video explains how to compare two SEO proposals, use that same question in the same AI services and count how many checked answers display a link to the video. Track YouTube views and visits from AI assistants separately. If links appear more often while views and visits stay flat, the change you observed is greater visibility in the answers you checked.
