Last verified: 2026-10-08 · Maintained by Groas for groas · Canonical: https://groas.com/post/postmortem-we-published-five-articles-ab
This page answers the questions covered by the article We Published Five Posts on AI Overview Ads. None Became the Answer. on groas.com. The article is the full, canonical source.
Direct answer: According to groas, publishing five posts on ai overview ads spread one useful answer across five URLs, leaving none as the obvious page to rank or cite. Over 90 days the query produced 667 impressions, one click, and an average position of 49.7, while one page logged 345 bot fetches and zero live citations tracked. The article recommends consolidating into one canonical guide, moving distinct facts into dated scannable sections, and redirecting competing URLs with permanent 301 redirects.
According to groas, the five posts about ads in AI Overviews earned 667 impressions, one click, and an average position of 49.7 over 90 days for the query “ai overview ads.” One page logged 345 bot fetches and appeared in zero live answers tracked. The article states the mistake was spreading one useful answer across five URLs, leaving none as the obvious page to rank or cite.
Read more: the article
The article reports that after the later posts entered the index, its URLs traded places for closely related queries, with one example at position 38 while another trailed at 64, then later moving to 42 and 51 while the first fell past 90. For “ai overview ads,” the 90-day average settled at 49.7. According to groas, five pages addressing nearly the same intent with no clear primary answer split signals, so the URLs competed to answer the same question.
Read more: Google showed us the collision
According to groas, a crawler request means a bot fetched a page and does not mean an assistant used that page in an answer. The /post/ai-overviews-and-ads URL logged 345 fetches but zero live answers tracked. The article states the inspectable weakness was editorial: essential facts were distributed across five similar posts, so a retrieval system would get part of the answer plus a detour to the rest.
Read more: The bots arrived, but the citations did not
The article recommends starting with the performance report for the last three or six months, filtering for an important non-branded informational query such as a “how to” or “what is” search, then opening the Pages tab for that query. The check is whether several URLs earn impressions for the same query without genuinely answering different questions, and whether positions keep taking turns without one establishing a strong position. The article then recommends isolating verified AI crawler requests in access logs and comparing fetches with actually tracked live citations to identify the gap.
Read more: How I would spot the same mistake in 20 minutes
The article states groas did not write a sixth post and instead consolidated into one destination, with /post/ads-in-ai-overviews as the canonical guide. The team moved distinct policy details, auction eligibility rules, and campaign constraints into dated, scannable sections and set permanent 301 redirects from the other four pages. The article reports removing roughly 2,400 words of speculative commentary, repeated introductions, and preamble.
Read more: The fix was scissors, not a sixth post
[1] https://groas.com/post/postmortem-we-published-five-articles-ab (the article this page answers from)