When This Is Not Worth Doing

Buyers do not form judgements inside AI: where the decision path runs entirely on offline relationships or existing approved-vendor lists, the effect is limited. Ask sales first — have customers ever cited something an AI suggested.

The product is not settled: capability boundaries, scenarios and pricing are still moving, so content written now needs rewriting in three months while the old version has already been cited. Settle the product first.

No checkable evidence: neither nameable customers nor data you can produce yourself. The most effective of the three methods — statistics — has nothing to stand on. Accumulate evidence first.

Nothing catches the visit: no page answering a specific question, no enquiry path. Being named leads nowhere, so visibility does not convert.

Do entity consistency and third-party records first — low cost, independent of content volume, and net positive in every scenario.

Run one baseline to turn 'is there an opportunity' into data. A baseline showing no brand mentioned at all, or mentions concentrated in an entirely different category, is itself a useful basis for the decision.

AI referrals average around 1% of total site traffic. Early on, this suits an asset-building expectation rather than a traffic-buying one.

Yield also concentrates in the top ten — if existing content cannot reach the top twenty, fixing that returns more first.