Notado vs. checking by hand
You asked ChatGPT once.
The same question, to the same engine, can get a different answer from one day to the next.
Example of 4 weeks of measurement, 3 answers per engine and week. ChatGPT: 15-09, named in 3 of 3 answers; 22-09, named in 3 of 3 answers; 29-09, named in 2 of 3 answers; 06-10, named in 3 of 3 answers. Perplexity: 15-09, not named in 3 answers; 22-09, not named in 3 answers; 29-09, not named in 3 answers; 06-10, named in 2 of 3 answers. Gemini: 15-09, not named in 3 answers; 22-09, not named in 3 answers; 29-09, named in 1 of 3 answers; 06-10, named in 1 of 3 answers. Claude: 15-09, not measured yet; 22-09, not measured yet; 29-09, not named in 3 answers; 06-10, not named in 3 answers. Copilot: 15-09, not named in 3 answers; 22-09, not named in 3 answers; 29-09, not named in 3 answers; 06-10, not named in 3 answers. AI Overviews: 15-09, named in 3 of 3 answers; 22-09, named in 2 of 3 answers; 29-09, named in 3 of 3 answers; 06-10, named in 3 of 3 answers. AI Mode: 15-09, named in 1 of 3 answers; 22-09, named in 1 of 3 answers; 29-09, named in 2 of 3 answers; 06-10, named in 2 of 3 answers.
We measure what these AI engines answer, in European Portuguese and in English.
On paid plans
- ChatGPT
- Perplexity
- AI Overviews
- AI Mode
- Gemini
- Claude
- Copilot
Unstable answer
The same question, minutes later.
- Three answers within one minute.
- That is why, in the full report, we repeat each question 3 times on each engine.
- We record every check, with its date.
Side by side
One question is not a check.
| Checking by hand | Notado check | |
|---|---|---|
| Engines tested | 1, whichever you have open | Free: ChatGPT and AI Overviews. Arranque (the full report, the fixes and a new measurement, paid once), 7: ChatGPT, Perplexity, Gemini, Claude, Copilot, AI Overviews and AI Mode |
| Questions | The one you think of typing | Tested by sector and city in Portugal, the way a customer would ask |
| Repeats per question | 1 run, 1 moment | Free: once. Arranque: 3 times on each engine |
| History | No record of what changed | We record every check, to compare with the previous one |
| Price | Free | Free, or €199 + VAT with the fixes and the new measurement |
One engine
One engine is not the market.
Your customer may use Perplexity, and the test you ran was on ChatGPT.
- BragaTest run on 1 September 2026.
- Four clinics per engineThe names each engine gave.
- One in commonAcross the engines tested.
The method
The same method in every check.
Measurement method
- Method version
- 20 November 2026
- Engines tested
- ChatGPT Perplexity Gemini Claude Copilot AI Overviews AI Mode
- Questions per site
- 107 on AI Overviews
- Repeats per question
- 3
- Answers read per check
- 2016 in the free check
- Latest calibration against the consumer product
- never calibrated
About the method
Why does the same question give different answers on different days?
Ask ChatGPT “which is the best dental clinic in Lisbon” today, and ask again tomorrow: the answers may not match. That is why the method repeats each question 3 times on each engine. A check by hand, done once, does not show whether what you saw is the rule or the exception.
Free check · no sign-up