What the machines say about your business. Fifteen minutes, once a month, same questions every time.
Write them the way a customer talks, not the way you would describe your own company. Casual phrasing finds errors that careful phrasing hides.
Four checks per answer, yes or no: NAMED = it mentioned you at all · FACTS = details right, starting with the owner · CITED = it referenced your actual website · LISTED = you appeared on the best-of list (Q4 only)
| Question | Assistant 1: ______ | Assistant 2: ______ | Worst error in the answer |
|---|---|---|---|
| 1 · Ownership | NAMED ☐ FACTS ☐ CITED ☐ | NAMED ☐ FACTS ☐ CITED ☐ | |
| 2 · What you sell | NAMED ☐ FACTS ☐ CITED ☐ | NAMED ☐ FACTS ☐ CITED ☐ | |
| 3 · Reputation | NAMED ☐ FACTS ☐ CITED ☐ | NAMED ☐ FACTS ☐ CITED ☐ | |
| 4 · Best-of list | LISTED ☐ FACTS ☐ CITED ☐ | LISTED ☐ FACTS ☐ CITED ☐ | |
| 5 · Recency | NAMED ☐ FACTS ☐ CITED ☐ | NAMED ☐ FACTS ☐ CITED ☐ |
Every wrong answer traces to something a machine read, and most of it sits on pages you control: an old press quote, a homepage with almost no readable text, an empty description field, a half-finished Yelp or Google profile. Write the fix next to the error and rank by damage.
| Error the machines made | Where it likely comes from / the fix | Done |
|---|---|---|
Same sheet, same wording, every month. Count the yes-marks per run (15 possible per assistant with Q4's LISTED). The trend is the result.
| Date | Assistant 1 score | Assistant 2 score | What changed since last run |
|---|---|---|---|