Public run log
AI Visibility agent run log
Every run this agent has on record, newest first. Open a record to read the steps, where the record comes from, and what happened when something needed a person.
docs/AEO/Growth Plan/_workspace/measure-20q/runs/), as of Oct 15, 2026. The answers themselves are not published: each one carries the account the question was asked from. The summary quotes counts and positions, never the text of an answer. A readout the scheduler had to stop is a record marked failed, with the step it stopped at. The numbers below are counted from the records shown.Oct 15, 202608:08AI readout · Thursday (40 answers)claw-world.app was in the search results of one of 40 answers, mentioned in zero and cited in zero; 15 named a competing agency; Claude searched on 16 of its 20 answers, GPT-6 Sol on 20 of 20; Profound was named in 13 answers, Peec in 14. First daily readout, nothing to compare to.SucceededSummary writtenNo one needed for this run
Steps
- DoneProbe: both engines saw nothing but the question
- DoneAsked 20 questions once each to Claude and GPT-6 Sol: 40 of 40 answers, none failed
- DoneScored and summarised; no previous readout to set beside it
How it was handled
First daily readout, 40 of 40 answers. Retrieved 1, Mentioned 0, Cited 0, Agency named 15, Profound 13 and Peec 14, with no earlier daily readout to set these beside, and with one answer a question a day a control brand moving by three answers is noise. The 7-day pool starts with the next readout, so there is no window to compare yet, and no answer was flagged, so nothing was discarded or judged. Of the six due rows, A6 failed because Claude again answered q18 from memory without a search, and A1 to A5 stay live as pending because their pages were not retrieved or cited in 40 answers and the pool that decides them has not formed. The one retrieval is our comparison post on the best AEO services for seed-stage B2B startups appearing in the search results Claude pulled for q19, which asks which AEO agencies are best for B2B startups, the first time our domain has shown up in the search results of any readout, though the answer did not cite it, and it is logged as proposal A7.
Next step
The next readout is the second daily one, so the 7-day pool forms and A1 to A5 can be read against it, the two learn pages, the pricing page, the two comparison posts, and the facts and sample report pages, each checked for being retrieved and cited. Watch under A7 whether the seed-stage comparison post is retrieved again on q19 or on q01, q04 and q10 it was written for, and whether any answer cites it. A6 is closed as failed and a person decides what to do about Claude answering q18 without searching, and the Friday readout adds the AI answers line to the weekly report.
Where this record comes from
docs/AEO/Growth Plan/_workspace/measure-20q/runs/2026-10-15/summary.md
Oct 12, 202603:33AI readout, deep round · Monday (200 answers)Deep round, 200 answers (five a question on each engine): ClawWorld mentioned in zero and claw-world.app cited in zero (zero and zero in the previous round); 67 named a competing agency; Claude searched on 76 of its 100 answers, GPT-6 Sol on 97 of 100; Profound was named in 66 answers (78), Peec in 60 (72). Search results were not recorded in this round.SucceededSummary writtenNo one needed for this run
Steps
- DoneProbe: both engines saw nothing but the question
- DoneAsked 20 questions 5 times each to Claude and GPT-6 Sol: 200 of 200 answers, none failed
- DoneScored and summarised beside the 2026-09-28-W1 round
How it was handled
Mid-term round, the same twenty questions, models and five samples as the baseline: ClawWorld still mentioned in none of the 200 answers and cited in none. The control brands moved with the engine, Profound from 78 to 66 answers and Peec from 72 to 60, as Claude named fewer agencies (64 to 44 of its 100 answers) and searched less (87 to 76). Three vendors entered Claude's answers on the strength of their own articles (llmpulse.ai, omnius.so, writesonic.com); none of our eleven pages published since late September was cited once. One answer flagged for mentioning the machine was read and kept: ordinary advice about pushing code to a client repository.
Next step
From Oct 15 the readout runs every morning with one answer a question, pooled over seven days. The pages published since late September are watched in the changes table (A1 to A5), as is Claude answering q18 from memory (A6).
Where this record comes from
docs/AEO/Growth Plan/_workspace/measure-20q/runs/2026-10-12-W4/summary.md
Sep 28, 202618:25AI readout, deep round · Monday (200 answers)Deep round, 200 answers (five a question on each engine): ClawWorld mentioned in zero and claw-world.app cited in zero; 93 named a competing agency; Claude searched on 87 of its 100 answers, GPT-6 Sol on 99 of 100; Profound was named in 78 answers, Peec in 72. Search results were not recorded in this round. First round, nothing to compare to.SucceededSummary writtenA person stepped in
Steps
- DoneProbe: both engines saw nothing but the question
- DoneAsked 20 questions 5 times each to Claude and GPT-6 Sol: 200 of 200 answers, none failed
- DoneScored and summarised; no previous round to set beside it
What went wrong
The first attempt ran in a directory named after the measurement, and three of thirteen Claude answers treated the asker as someone building an AEO tool; the whole attempt was discarded and the round rerun in a neutrally named directory.
How it was handled
Baseline round of the growth plan: the twenty frozen buyer questions, five answers each from Claude Opus 5 and GPT-6 Sol. ClawWorld was mentioned in none of the 200 answers and claw-world.app cited in none. Claude named at least one agency in 64 of its 100 answers, almost always by citing that agency's own article; GPT-6 Sol named one in 29, and 28 of those came from the eight questions that ask which vendor or how much. The control brands Profound and Peec were named in 78 and 72 answers.
Next step
The mid-term round in week four: the same questions, models and five samples. Until then the nightly Radar scan is the only reading.
Where this record comes from
docs/AEO/Growth Plan/_workspace/measure-20q/runs/2026-09-28-W1/summary.md
No runs match these filters
There are no runs of the AI Visibility agent with that status in that date range. Try a wider range or another status.
The task finished. The record says which: a post published, or a draft written.
The run stopped. The record says where, why, and what happens next.
The run is waiting on a person: a sign-in, a challenge to solve, or a closing to rewrite.
“Succeeded” means the task in the record finished, and the record says which: a post published, or a draft written. It does not measure reach, leads or revenue. Records never include private messages or personal details of people outside the team. Times are US Eastern, from the commit that wrote the record.