AIOuditJournal
← All reports

Observation note

Checking AI search visibility while logged in gives you the wrong answer

On October 26, 2026, we asked ChatGPT about the same topic twice. In a regular chat on the account we use every day, the answer brought up our own service by name. In a Temporary Chat, which uses no history, it did not. What your customers see is much closer to the second answer.

October 26, 2026CHROMETI Research4 min read

What we observed

Both questions were on the same topic: recommend GEO audit services that can be used in Korea. The second question also asked for YouTube audit tools, and the wording differed slightly. GEO (generative engine optimization) means shaping your pages so that they appear as sources inside AI answers.

In a regular chat on the account we use for work, ChatGPT introduced several services and then, without being asked, named our own service and continued along the lines of “if you want to validate a service like this.” When we asked again the same day in a Temporary Chat, it recommended only other services. Our name did not appear once.

Regular chat · uses account history Our name appears though the question never mentioned it Earlier conversations on the account appear to have leaked into the answer. This does not happen on a customer’s screen.
Temporary Chat · no history Only other services recommended, our name absent Much closer to what a customer asking this for the first time receives. This is the setup to measure with.
Same day, same topic, different answers. ChatGPT, observed October 26, 2026. Answers are shown in outline only.

Why the answers differ

ChatGPT can personalize answers using memory and past conversations. According to OpenAI’s help center, a Temporary Chat starts from a blank slate without past conversations or saved memories, and it does not appear in your chat history.

So if you check “do we show up in AI?” from an account that has talked about your brand many times, the AI may mix that context into its answer. The result looks better than reality and gives you false reassurance.

One more thing. The same help article notes that custom instructions still apply in a Temporary Chat if you have them turned on. Turn them off before you measure.

How to check accurately

  1. ChatGPT Ask in a Temporary Chat Turn off custom instructions, switch on Temporary Chat, then ask.
  2. Gemini Use a temporary chat in Gemini too Gemini’s temporary chats are not used to personalize your experience.
  3. Record Write down the date and conditions Keep the exact question, date and model so you can compare next month.

The same goes for Google. Temporary chats in the Gemini app are reported to stay out of your history, are not used for personalization or model training, and are kept for up to 72 hours so Gemini can respond and process feedback.

Takeaway

The AI answer you see on your own account is your answer. To know what customers see, ask with history turned off.

From now on, AIOudit Journal will make every observation this way and publish each result with its date and conditions.

Sources

  1. OpenAI Help Center · Temporary Chat FAQ
  2. 9to5Google · Report on Gemini temporary chats (Aug 13, 2025)
  3. CHROMETI Research · Direct observation in ChatGPT (Oct 26, 2026)