Why this question is suddenly everywhere
Your client's customers changed how they shop for answers. When someone asks ChatGPT "best CRM for a 10-person agency," the reply names three or four brands — and if your client isn't one of them, that's a shortlist they never knew they missed. No impression data, no referral string, no consolation-prize ranking on page two. Just absence.
The uncomfortable part for agencies: clients assume you already know what the engines say. You do SEO, this looks adjacent, therefore it's your department now. Congratulations on the promotion.
The four buckets
Every engine answer about a brand lands in one of four buckets, and each bucket has a different fix. This is the sorting that turns a vague worry into a work plan.
- Present and accurate. The engine names the client and gets the facts right. Your job is defense: keep the sources fresh so it stays true.
- Present but wrong. Named, but with stale pricing, a dead product line, or a description from two rebrands ago. Usually a training-data echo — fixable, but through the sources engines read, not through wishing.
- Absent. Competitors get named; your client doesn't. Almost always a footprint problem: the comparison posts and review roundups engines lean on simply don't include them.
- Misattributed. The engine blends your client with a similarly named company. Entity ambiguity — annoying, common, and very solvable with consistency work.
Before the meeting: a 30-minute sweep of your own
Ask ChatGPT, Gemini, Perplexity, and Claude the three or four questions the client's actual customers would ask. Not "tell me about Acme" — nobody shops that way. Ask category questions: "best X for Y," "alternatives to [market leader]," "is X worth it." Record who gets named, in what order, and what gets claimed.
Two things to watch: whether answers carry citations (that's live retrieval — you can trace the source) and whether they don't (that's memory — a longer game). The split tells you which fixes are fast and which are quarterly.
After the meeting: promise measurement, not magic
The honest pitch is that AI answers can be moved — engines ground answers in sources, sources can be improved — but the timeline is weeks for retrieval-driven answers and quarters for memorized ones. Promise a baseline, a monthly re-check, and movement on specific answers. Don't promise "we'll get you into ChatGPT" as if there's a form to fill in. There isn't, and your client will eventually ask to see it.