Why it matters.
Google’s own documentation on AI features says that “AI Mode and AI Overviews may use different models and techniques, so the set of responses and links they show will vary.” If two Google products differ, ChatGPT and Perplexity differ more, because each one searches its own index or partner index and reads different pages.
A firm can be named often by Perplexity and rarely by ChatGPT. One blended score would call that “average” and point you at the wrong fix.
The 10-minute self-check.
Draw a table with one row per engine and these four columns, then fill it from your last runs:
- Named rate: share of runs naming you.
- Sources: the three domains it cites most for your questions.
- Accuracy: asked about you by name, does it get your services, location and founding year right?
- Competitors: the names it gives most often instead of you.
Any empty cell is the first thing to measure.
Steps.
- Choose the engines your buyers use. For most B2B firms: ChatGPT, Perplexity, Google AI Overviews and AI Mode.
- Run your fixed panel (play 02) the agreed number of times per engine (play 03).
- Fill the four columns per engine. Keep the raw answers so the scores can be checked.
- Mark the weakest column per engine. That decides the next play: sources point to category 4, accuracy to category 2.
- Report each engine separately. If leadership wants one number, show the range, not an average.
Our AI search visibility scorecard uses these four columns.
What we saw on prolinkage.com.
How to measure it.
- Per-engine named rate: monthly, same panel.
- Accuracy errors: count of wrong facts per engine, by name. Monthly until zero.
- Source overlap: how many of each engine’s top sources mention you. Quarterly.
Sources.
- Google Search Central, AI features and your website, read 8 November 2026.
- Prolinkage AI self-test (ChatGPT, Perplexity, Google AI Overview, one run each), 6 November 2026.
- Prolinkage change log: Clutch and DesignRush profile corrections, 7 November 2026.