GPT Visibility agent run log

How it was handled

Against the round of 24 October: none of the 19 category queries held one of our pages (0 of 19), GPT-6 Sol's search can find 19 of our 23 target pages (19), 2 of the 19 titles it holds are out of date (2), a search for our own name returns us 8 times in 15 results with the first at position 1 (8 of 14), and in the six daily AI readouts from 19 to 24 October its own searches returned one of our pages in 0 of 120 answers (0 of 140 over seven readouts). Nothing moved among the target pages or the diagnoses: 15 queries are indexed but not ranked, 3 have no page written for them, 1, the Google AI features guidance query, still waits on /learn, and the q10 query still does not list /blog/best-aeo-services-seed-stage-b2b-startups-2026, revised for it yesterday; two pages that are not targets entered the index GPT searches, /learn/aeo-vs-seo and /blog/aeo-agency-pricing-statistics-2026-done-for-you-monthly-cost, and the two stale titles are yesterday's, /blog/how-to-tell-whether-aeo-worked still shown with the title it had before 20 October and /vs with the one it had before 5 October. The four target pages the search cannot find, /pricing, /learn, /learn/ai-crawlers-no-javascript and /audit, went to Bing's URL submission API this morning, logged as G5 with a check on 22 November, which closes the proposal of 18 October that asked for it, G2, as superseded. No row came due and no proposal was added, proposals being a Friday job; 432 Reddit threads went to the comment agent's list, 51 of them holding three or more of the buyer questions and none of them commented under yet.

Next step

Two rows come due in the next seven days: G4, the revision of /blog/best-aeo-services-seed-stage-b2b-startups-2026 for the q10 query (still found, listed by that query, returned by GPT-6 Sol's own searches in the AI readouts), on 31 October, and G1, the watch on the stale title of /vs, on 1 November. G5, the Bing submission of /pricing, /learn, /learn/ai-crawlers-no-javascript and /audit, is checked on 22 November, and every round before then should say whether any of the four entered the index GPT searches; the one open proposal, G3, is seven days old and not yet waiting on a person. The Friday round of 30 October adds the week's proposals and the weekly report line, and should say whether the snapshot of /blog/how-to-tell-whether-aeo-worked has caught up with its new title.

Where this record comes from

docs/AEO/Growth Plan/_workspace/gpt-serp/runs/2026-10-25/summary.md

How it was handled

Against the incomplete round of 18 October: none of the 19 category queries held one of our pages (0 of the 6 read then), GPT-6 Sol's search can find 19 of our 23 target pages (18), 2 of the 19 titles it holds are out of date (1 of 18), a search for our own name returns us 8 times in 14 results with the first at position 1 (6 of 21), and in the seven daily AI readouts from 18 to 24 October its own searches returned one of our pages in 0 of 140 answers (0 of 80 over four readouts). /local-businesses entered the index GPT searches, which moves the two pricing queries from awaiting index to indexed, not ranked; with every query read for the first time, 15 are indexed but not ranked, 3 have no page written for them (OpenAI's and Perplexity's crawler documentation, and a controlled study of generative engine optimization), and 1, the Google AI features guidance query, waits on /learn, which the search still cannot find. The new stale title is /blog/how-to-tell-whether-aeo-worked, still shown as How to Tell Whether AEO Worked, the title it carried before the post was updated on 20 October, and /vs is still held with its title from before 5 October. The four target pages the search cannot find, /pricing, /learn, /learn/ai-crawlers-no-javascript and /audit, are due for the Bing submission experiment, which did not run this morning; no row of the table came due, and no proposal was added, proposals being a Friday job. 495 Reddit threads went to the comment agent's list, 53 of them holding three or more of the buyer questions and none of them commented under yet.

Next step

No row comes due in the next seven days: the first is the watch on the stale title of /vs, on 1 November, and the two open proposals of 18 October (the Bing submission of the pages GPT's search cannot find, and the Reddit thread list for the comment agent) are six days old, so neither is flagged as waiting on a person yet. The next round should say whether /pricing, /learn, /learn/ai-crawlers-no-javascript and /audit went to Bing, and whether the snapshot of /blog/how-to-tell-whether-aeo-worked has caught up with its new title; the Friday round of 30 October adds the week's proposals and the weekly report line.

Where this record comes from

docs/AEO/Growth Plan/_workspace/gpt-serp/runs/2026-10-24/summary.md

What went wrong

The round came back incomplete: 14 of the 44 planned probes returned no result list, so only 6 of the 19 category queries have a reading and the first number is counted out of 6 rather than 19. Nothing was discarded and no probe carried an isolation warning.

How it was handled

This is the first round, so there is nothing to set the numbers beside: none of the 6 category queries that came back held one of our pages, GPT-6 Sol's search can find 18 of our 23 target pages, 1 of the 18 titles it holds is out of date, a search for our own name returns us 6 times in 21 results with the first at position 1, and across the four daily AI readouts of this week its own searches returned one of our pages in 0 of 80 answers. Of the six queries read, four already have a page of ours that GPT's search can find and simply do not list it, and two are waiting on the index: the pages written for the two pricing questions, /pricing and /local-businesses, cannot be found at all. Five target pages are invisible to that search while the rest are not, /pricing, /learn, /learn/ai-crawlers-no-javascript, /audit and /local-businesses, and all five are due for the Bing submission experiment, which this round did not run. The comparison page /vs is held with the title it carried before 5 October; that watch is already on the table and comes due on 1 November. No check was due and no proposal was added, proposals being a Friday job, and 377 Reddit threads went to the comment agent's list, 28 of them holding three or more of the buyer questions and none of them commented under yet.

Next step

No row comes due in the next seven days: the watch on the stale title of /vs is set for 1 November, and the two open proposals, handing the five pages GPT's search cannot find to Bing's submission API and reading the Reddit thread list into the comment agent, are both waiting on the first scheduled run rather than on a person. The next round should get a result list for the 13 category and brand queries that returned none this time, so the first number is counted out of 19 again, and should say whether the five missing pages were submitted.

Where this record comes from

docs/AEO/Growth Plan/_workspace/gpt-serp/runs/2026-10-18/summary.md