Short answer: being recommended requires three things at once — the engine can reach you, it can extract a claim about you, and something it treats as independent corroborates that claim. Most brands have the first, half of the second, and none of the third.
Three outcomes, not one
The vocabulary matters here because the fixes differ:
| Outcome | What it means | What moves it |
|---|---|---|
| Mentioned | Your name appears in the answer | Entity clarity, third-party presence |
| Cited | A URL of yours is listed as a source | Crawl access, extractable passages |
| Recommended | The answer actively suggests you | Corroborated, comparable claims |
A brand can be cited without being recommended — the engine used your page as reference material and then suggested someone else. It can also be recommended without a single link to it. Reporting these as one number hides exactly the gap you need to see.
Why corroboration is the bottleneck
An engine writing a recommendation is making a claim it has to stand behind. Claims sourced only from the vendor's own marketing page are the weakest kind. What changes the outcome is the same claim appearing somewhere the model treats as independent: a comparison page that isn't yours, a community thread, a directory listing with real fields filled in, documentation written by a customer.
You cannot write your way into being recommended using only your own domain. That is not an SEO rule; it is how the model weighs evidence.
Make your claims comparable
Recommendations happen inside comparisons. "Best tool for X" forces the engine to line up candidates on shared dimensions — price, coverage, who it's for. If your page states none of those in extractable form, you are not a candidate, you are background reading. State the dimensions plainly, including the ones where you lose. A page that admits a limitation is more usable to a model constructing a comparison than one that claims to be best at everything.
Then measure the two separately
Track mention rate and recommendation rate as separate columns over a fixed question set. A brand mentioned in 40% of answers and recommended in 4% has a corroboration problem. A brand with the reverse pattern has an attribution problem. One number cannot tell you which you have.