Why does Outercite show fewer citations than a manual search?
A text search counts every appearance of your name. Verification counts recommendations. The gap is the whole point of the number.
If you ask an engine yourself and see your name, then look here and see a lower rate, nothing is broken. The two are counting different things on purpose.
What you'll learn
- What a raw search counts
- What is deliberately excluded here
- Why the smaller number is the useful one
What a manual check counts
Every appearance of your name. That includes:
- a passing reference with no recommendation attached
- your name inside a list of everyone in the category
- a generic term that happens to match your business name
- an answer naming you as the thing to avoid
- your own marketing being quoted back at you from a page you wrote
All of it counts as a mention. None of it necessarily means the engine would recommend you.
What is counted here
Your business presented as a genuine option. Everything below is excluded, deliberately:
| Excluded | Why |
|---|---|
| A generic term matching your name | "GRP tubing" the material is not "GRP Tubing Pty Ltd" the company |
| A negative recommendation | Being named as a warning is not a citation. It is recorded as a negative mention |
| A citation the two models disagree about | Disagreements fail closed |
| A check with no signal at all | Nothing to judge |
| An absent AI Overview | Google did not draw a panel, so there was no answer to be cited in |
Why the smaller number is the useful one
Three reasons:
- It is decision-grade. A count that includes "avoid this company" is a count you cannot use.
- It is comparable over time. A strict definition applied consistently produces a trend you can act on. A loose one produces noise that moves with how chatty the engines were this week.
- It survives scrutiny. You can open any counted citation and read the sentence. That is not true of a number built from string matching.
The trade you are making
The strict definition will occasionally exclude something you would personally have counted. That is the cost of a number that is wrong in the direction of understating you rather than flattering you.
If you want to see the borderline cases, they are all readable: low-confidence rows in Proof, and negative mentions in their own view.
Try this in Outercite
Ask the assistant "Show me the exact sentences AI wrote about us", then run the same prompt yourself in an engine. The difference between what you counted and what the platform counted is the definition doing its job.
Related
How verification works
The pipeline that applies the definition.
What is a citation?
The five forms that do count.
How a citation is counted
Confidence and consensus, explained.
Last updated on