For agencies: adding AI answer visibility to a local SEO retainer

This one is for agencies and freelance consultants who look after local businesses: trades, clinics, restaurants, salons, law firms. Your clients have started asking "are we showing up in ChatGPT?" Some have tried it themselves and been either delighted or alarmed by a single answer. This article covers how to turn that question into a piece of work you can sell, deliver and report on without overclaiming.

Why it fits a local retainer

Most of what improves a local business's chances of being named in AI answers is work a local SEO agency already does:

• a complete, accurate Google Business Profile;
• consistent name, address, phone and hours across the listings that matter for the trade;
• reviews, and replies to them;
• a website that states clearly what the business does and where, readable by search crawlers;
• structured data that matches the page.

AI answers don't replace that work. They change how its results are seen. A search results page shows ten businesses and a map; an AI answer often names three. So the question a client cares about becomes sharper: are we on the shortlist, and who is on it instead of us? That's a question you can measure and report on every month, which is what makes it a retainer line rather than a one-off project.

Step 1: the pitch audit

Prospects respond to their own name and their competitors' names far more than to charts about the industry. A good AI visibility audit for a prospect answers three things on one page:

1. How often are they named? Across the main engines, for the questions their customers actually ask, compared with three local competitors the prospect will recognise.
2. Where do the AIs get their information? The listing and review sites cited in those answers, and which of them mention the competitors but not the prospect.
3. What would you do first? Three concrete fixes, each tied to something you found.

Keep it short and specific. "Named in 2 of 30 answers; Smith & Sons named in 19; Perplexity cites Checkatrade for most of these questions and you're not listed there" is a better opening line than any slide about the future of search.

Use the prospect's real competitors. Pick businesses of the same type, close by, that the owner already worries about. If the competitors are wrong, the whole audit reads as generic.

Step 2: measure honestly

This is where most AI visibility reporting goes wrong, and where an agency can set itself apart.

AI answers vary. Ask the same question twice and you'll often get a different list. The consumer apps can personalise answers, so what your client sees on their own phone isn't what a stranger sees. We go into both in Why ChatGPT tells you something different from what your customers see. Two consequences for reporting:

Report counts with a range, not a single score. "Named in 11 of 40 answers (28%)" tells the client what happened. Adding the likely range, here about 18% to 40% at 90% confidence, tells them how sure that figure is. The range comes from a standard statistical method for proportions (we use the Wilson score interval (external site)), and it narrows as you collect more answers. A score of "64.3 out of 100" from a dozen answers implies a precision that isn't there.

Only call a change when it's bigger than the noise. If last month's range was 18% to 40% and this month's is 22% to 45%, the honest summary is "about the same". If the ranges no longer overlap, something has really moved. This protects you as much as the client: you won't be celebrating a lucky month and then explaining a "drop" the month after that was never real.

Two more rules worth adopting:

• Missing is not "not named". If an engine errored or didn't answer, leave that answer out of the figure. Counting it as a miss makes clients look worse for reasons that have nothing to do with them.
• Say how each answer was collected. Whether through an official API with a location set, or from Google's results pages, the method should be visible. API answers approximate the consumer apps; they don't reproduce them. Clients deserve to know that.

Step 3: report in the client's language

Your client doesn't want an AI dashboard. They want to know whether it's working and what you did. A monthly AI visibility section that works for non-technical owners has four parts:

1. One headline sentence. "This month AI answers named you in 14 of 60 (likely 16% to 33%), about the same as last month. Smith & Sons: 28 of 60."
2. What changed, in plain words. Only changes that clear the noise. "ChatGPT now names you for 'boiler repair Headingley'; it didn't in September."
3. What we did. The fixes completed, each with a note that you checked it went live: "Corrected your hours on Yelp (checked 4 October)."
4. What's next. The next two or three fixes, and why.

When a fix and an improvement coincide, say "possibly related", not "caused by". Many things change between two runs, including the engines themselves. Clients trust an agency that doesn't claim every upturn.

What not to promise

• No rankings guarantees. Google's own local ranking page (external site) says there's no way to request or pay for a better local ranking. Nobody can guarantee a place in an AI answer either; the vendors don't publish how names are chosen.
• No "we'll get you into ChatGPT in 30 days". Some changes show up within weeks, especially fixing blocked crawlers or missing listings. Others take longer or don't move the figure at all.
• No promises about what one person sees. Because of personalisation and run-to-run variation, a client will sometimes check on their phone and not see themselves after a good month. Set that expectation in the first meeting, and show them why your measured figure is the better guide.

Pricing the work

How you price it is your call. Three approaches fit the work well: a paid or free audit as a lead generator, a monthly add-on to existing local SEO retainers, or folding it into a higher tier. Whichever you choose, make the tracking cost per client small and predictable, so the margin on the add-on is clear from the first month.

How Saymeter supports this

Saymeter's agency plan is built around this workflow: a table of every client with their share of AI answers and trend, branded pitch audits from a prospect's name in a couple of minutes, client view-only access, and reports under your own brand. Every figure is shown as "named in X of Y" with a likely range, and changes are only called when the ranges stop overlapping. How we collect and read answers is on our methodology page. You can try it on any business with a free scan.