Did it work? Marking a fix done and seeing the result
Press Mark done on an item in your to-do list. Your next checks of the same brand and market measure it again and say whether it worked. What the AI assistants say is judged from 14 days after you mark it.
3 min read · Updated 5 November 2026
How it works
- 1In What to do next, your to-do list ranks what the check found, most impact first.
- 2When you have done one, press Mark done beside it. PollyX keeps that item's numbers from this check, for example "named first in 0 of 11 answers", and shows the date it will be judged from.
- 3Your next checks of the same brand and market measure the same numbers again. The Did it work? card shows each fix with its result and its numbers then and now.
You need a later check to see a result: run one, or track the brand so checks run by themselves.
The results
- Worked: the numbers moved the right way, and enough time has passed to credit your fix.
- Moving the right way: they moved the right way, but it is still under 14 days since you marked it. A good sign, not proof yet.
- Too early to tell: no change yet, and it is still under 14 days.
- Not yet: 14 days or more have passed and the numbers have not moved. Look again at what you changed, or try another fix.
- Waiting: there is no later check yet.
- Cannot measure: the later check did not ask that question or run that test.
Why AI answers wait 14 days
AI assistants do not give the same answer twice. Ask one the same question tomorrow and it may name different brands even if nothing changed: we have seen a question go from first in 0 of 10 answers to first in 3 of 11 in one day, with no fix at all. And when you publish a page, the assistants need time to find it and read it.
So for anything about what the assistants say (a question, a site they read, a fact they state), PollyX credits your fix only from 14 days after you marked it. Before that, a gain shows as Moving the right way and no change as Too early to tell.
What counts at once
Two kinds of fix are exact, so your next check settles them straight away:
- a website test, for example your robots.txt no longer blocks the assistants;
- broken links: the pages now open. See Broken links the assistants send people to.
An example
Mary's to-do list says the assistants never name her coffee brand for "best coffee beans in Ireland". On 5 November she publishes a page that answers it and presses Mark done: first in 0 of 11 answers, judged from 19 November. Her check on 12 November shows first in 2 of 11: Moving the right way. Her check on 26 November shows first in 3 of 11: Worked.
Mark again and Undo
- Undo removes the mark.
- Mark again, on a later check, measures the fix from that check's numbers instead.
One check is a sample: a fix that worked can still wobble on the next check, so look at a few checks in a row. Only you see your marks; they are in your own downloads too (PDF, slides and Markdown), never on a shared link.
Was this article helpful?
Related articles
- Reading your AI Visibility resultsThe 0 to 100 score, the margin of error, the question-by-assistant matrix and the chapters of a completed check.
- Tracking your AI visibility over timeTrack this brand re-runs the same questions weekly, every 2 weeks or monthly, charts the movement and alerts you when your position shifts.
- Why you get a different answer when you ask an assistant yourselfAI assistants are not deterministic: the same question can get different wording, order and brands. A check measures the pattern across many answers.
- Broken links the assistants send people toPollyX opens the pages of your site that the AI assistants read and lists the broken ones, so a buyer who follows an AI answer does not land on an error.