Compare responses from the same service to the same question in two reports. If the question changed or no text was received in the second period, that pair cannot be recorded as a lost mention.
Suppose two services mention the company again, while a third returns no response. The third result is unknown. To explain the overall counter, also state the number of responses received and the dates of the checks.
Keep the exact wording
Before the next round of question edits, save their text in a working document. Give the versions clear dates and explain why a new set was introduced. This is an external way to organize observations, not a promise of built-in version management.
Even a small addition of a brand, city, or condition changes the task for the model. The questions “who services offices” and “what is known about company X” cannot be treated as identical just because both lines relate to the same service.
The guide to building an SEO semantic core offers a useful approach to grouping customer tasks. When comparing months, the specific wording must remain stable—not just the broad topic of the section.
Identify questions that did not change
If two of five lines were updated, the other three may still be suitable for comparison. Do not discard all useful history, but do not combine new questions with old ones as if they were one series.
Create a table with the question text, whether it appeared in the first report, whether it appeared in the second, and whether its conditions changed. Use only the clearly comparable portion to assess overall trends, and state the size of that sample.
A new question establishes a baseline without historical trends. A discontinued question describes an old task. It does not become a negative result for the current month just because it is no longer being asked.
Check service availability
reChecker supports ChatGPT, Alice, DeepSeek, Gemini, Claude, Grok, and Perplexity. However, the actual report may cover fewer than the expected seven services. Check the statuses and dates for each available result.
For a pair where a response was available before but data is now missing, record “unknown.” This is different from a response that was received but did not mention the brand. Substituting zero will exaggerate the decline and create a fictitious negative response.
Monitoring is scheduled monthly. If more than a month has passed between available reports, state both actual dates. Do not call the interval a calendar month just because that is the usual heading in the summary.
Map the transitions in a table
Take two sample reports for the same unchanged question. The first contains three responses; one service is unavailable for the second. Describe the transitions separately.
| Service | First report | Second report | Conclusion for the pair |
|---|---|---|---|
| Service A | Brand mentioned | Brand mentioned | Mention persisted |
| Service B | Brand not found | Brand mentioned | Mention appeared |
| Service C | Brand mentioned | No data | Change is unknown |
The totals from the first and second reports in this example do not tell the whole story. One positive transition is confirmed, while the outcome for another pair is undetermined.
For an external chart, label the number of comparable pairs. If coverage changed, make that visible alongside the result rather than hiding it in a note after the overall conclusion.
Read the context of a continuing mention
A brand that is still mentioned may now have a different role. Previously, it may have been named as a suitable provider; now it may be cited as an example of an outdated service. The same unit in the counter does not describe this difference.
Open both texts and check that the company was identified correctly. Compare the service, location, and material facts. If there is an error, call it out separately from the “name is present” transition.
OpenAI warns that responses may be inaccurate. For monthly work, this means manually checking important information rather than treating a confident-sounding text as confirmation of a change in the business.
Choose one pair to examine next
Choose one confirmed pair with a change and explain it in words. For example, the name appeared in a response to an unchanged question about a specific service; the statement matches the company; the link has been checked. This is more meaningful than a general claim that “AI visibility has increased.”
For a negative transition, first examine the text and public information. The reason may remain unknown. Do not add the brand to the question arbitrarily just to change the counter if the goal was to choose a provider neutrally.
Regular monitoring through the reChecker subscription helps build up observations. It does not promise that responses will remain constant or that mentions will increase after every revision. Any action taken should be useful to visitors and have a verifiable result on the website.
Read more: Quarterly website review in 60 minutes: agenda and minutes.