Recommendations, and what they are not
How Citerva turns scan evidence into things worth investigating, and why none of them is a promise.
After a scan finishes, Citerva reads what came back and writes a short list of things worth looking at. Every one of them points at evidence from that scan.
What a recommendation is
A recommendation is an observation plus a suggested place to look. It is not a prediction, and it is not a task list.
Citerva will not tell you that publishing a comparison page will make ChatGPT recommend you. Nobody can tell you that. What it can tell you is that you were absent from four prompts where a competitor was named, and that all four were comparison questions — which is a fact about your scan, and a reasonable place to spend an afternoon.
If you ever read a recommendation here that promises an outcome, it is a bug. Tell us.
The six kinds
| Kind | What triggered it |
|---|---|
| Content gap | Prompts where nobody named you, grouped by topic |
| Citation gap | Answers that cited sources, none of them yours |
| Competitor gap | Prompts where a specific competitor was named and you were not |
| Prompt coverage | Topics your approved prompt set does not ask about |
| Comparison content | Comparison-style prompts specifically, where absence tends to be most actionable |
| Provider gap | An assistant that names you far less often than the others |
Each one names the prompts it came from, so you can open the scan, read the answers, and decide for yourself whether the recommendation is worth anything. That is the intended workflow: the recommendation is the index, the evidence is the argument.
Working through the list
A recommendation has a status and a priority, and both are yours to set.
- Open — nobody has looked at it yet. This is where they start.
- Investigating — someone is on it. Useful when more than one person reads this list.
- Done — you have acted on it. It stays visible, struck through, because a recommendation you have addressed is part of the record of what you did.
- Dismissed — you have decided it is not worth acting on. This is the only status that removes it from the list.
Priority is high, medium or low. Citerva sets an initial priority from the size of the gap; you should override it whenever your judgement differs, because Citerva does not know what your business is trying to do this quarter.
Nothing here is destructive. A dismissed recommendation is marked, not deleted.
Why the list does not empty itself
Recommendations are generated per scan, from that scan's evidence. Acting on one does not remove it — the next scan is what tells you whether the action worked, by whether the same gap appears again.
That is deliberate, and it is the honest version. A list that cleared itself the moment you clicked "done" would be measuring your clicking, not your visibility.