Agency Workflow: When a Client Asks "Are We on ChatGPT?"

Agency workflow for AI visibility starts the moment a client sends that email: 'Hey, someone told me we don't show up in ChatGPT. Is that true? Should we care?' It's happening more often now. And the agencies that have a crisp answer — and a process behind it — are the ones turning that question into a new service line.

Step 1: The Discovery Call (Don't Wing This)

Before you touch any tools, have a 20-minute conversation. You need to understand what the client actually sells and who they sell it to. A personal injury law firm in Phoenix and a SaaS HR platform have completely different AI visibility problems. ChatGPT answers 'best HR software for small business' very differently than it answers 'car accident lawyer Phoenix.' One involves dense comparison tables. The other involves local authority signals. You need to know which fight you're in.

• —What search queries would their ideal customer type into ChatGPT or Perplexity?
• —Have they noticed competitors being recommended by name in AI answers?
• —Do they have existing structured content — FAQs, service pages, about pages?
• —Are they currently doing any traditional SEO, and does their agency know what AEO is?

Step 2: Run the Baseline Score

Pull their URL into Zygur. You get a 0–100 AI visibility score in about 60 seconds. This number becomes your anchor for every conversation that follows. Clients don't need to understand schema markup or llms.txt — they need a number. A 34 out of 100 is a story. It explains why Perplexity cites their competitor and not them. It explains why Claude returns a generic answer when someone asks for a recommendation in their category. The score is your opening slide and your closing argument.

Step 3: Interpret the Score Like a Human

Here's the counterintuitive part: a low score is actually the better sales situation. A client at 31 has clear, fixable problems. A client at 78 is harder to pitch because there's less obvious pain. The score breakdown tells you what's dragging them down — missing structured data, no FAQ schema, weak entity signals, no llms.txt file. These aren't abstract technical complaints. Translate them: 'ChatGPT can't find a clean answer to who you are and what you do, so it skips you and cites someone whose site makes it easy.'

Step 4: Show Them the Actual AI Output

Open ChatGPT, Perplexity, Claude, and Gemini in front of them — or record a Loom and send it. Type in the query their customer would actually use. Show what comes back. If a competitor is named and they aren't, that's the whole presentation. You don't need a deck. Perplexity in particular is useful here because it shows citations — you can literally point to the URLs it pulled from. If your client's site isn't there and a competitor's is, the visual does the work for you.

Step 5: Deliver a Fix Proposal (Not a Strategy Document)

Don't send a 14-page strategy PDF. Send a one-page fix list with three columns: what's broken, what the fix is, and what it affects. Agencies using Zygur can run automated fixes directly — the platform generates structured data, FAQ schema, and content recommendations without manual writeups. Your proposal is basically: here's your score, here's what's dragging it down, here's what we do in month one, here's what it should move to. Clean, fast, billable.

Step 6: Execute Month One

• —Deploy or fix structured data — Organization, LocalBusiness, Service, FAQ schema as applicable
• —Add or clean up an llms.txt file so AI crawlers have a direct map to key content
• —Audit and expand FAQ content — Gemini and ChatGPT both over-index on well-structured Q&A
• —Strengthen entity signals — NAP consistency, Wikipedia-style factual anchors, clear 'About' language
• —Rescore at end of month to document movement

Step 7: Build the Retainer Around Monitoring

This is where the service becomes sticky. AI search isn't static. ChatGPT's browsing behavior changes. Perplexity updates its ranking signals. New competitors build better structured content. A monthly retainer built around Zygur monitoring gives you a recurring reason to be in the client's inbox: 'Your score moved from 54 to 71. Here's what changed. Here's what we're watching next month.' That's a retainer that justifies itself every billing cycle without you writing a 10-page report.

What to Charge

Baseline audit: $500–$1,500 depending on site complexity. Month one implementation: $1,500–$3,000. Monthly monitoring and optimization retainer: $500–$1,500/month. These are conservative numbers. Agencies with established AEO practices are charging more, especially in competitive verticals like legal, finance, and healthcare where AI recommendations carry real referral weight. The pricing floor is whatever you'd charge for a technical SEO audit — AI visibility is the same conversation with a different audience.

Frequently Asked Questions

How do I know if a client's site is actually being recommended by ChatGPT?

Run the queries their customers would use in ChatGPT, Perplexity, Claude, and Gemini. Look for brand name mentions and URL citations. Perplexity and Gemini show sources directly. ChatGPT with browsing enabled will sometimes cite URLs. If the client's brand isn't appearing in 5–10 relevant queries, that's a measurable gap. Zygur's score also surfaces this as part of its visibility analysis.

Is AI visibility the same as SEO, or is this a different service?

Related but different. Traditional SEO optimizes for Google's ranking algorithm. AI visibility optimizes for how large language models read, parse, and cite content. Structured data matters in both, but AI systems weight FAQ schema, entity clarity, and llms.txt more heavily than Google does. You can have strong Google rankings and terrible AI visibility — it's common, actually.

What's llms.txt and why does it matter for AI visibility?

llms.txt is a plain-text file placed in a site's root directory that tells AI crawlers and agents what pages are most important and how to interpret the site. Think of it as robots.txt but written for language models instead of search bots. Sites without one leave it to the AI to guess what matters. Sites with a clean llms.txt file give AI systems a direct map. Claude and Perplexity both interact with it during crawls.

How long does it take to see movement in AI recommendations after making fixes?

Structured data and llms.txt changes can be picked up within days if an AI crawler revisits the site. Meaningful changes in how ChatGPT or Perplexity responds to queries typically show within 2–6 weeks, depending on recrawl frequency and how competitive the query is. This is why monthly monitoring matters — the feedback loop is slower than paid search but faster than most clients expect.

Can small agencies actually sell this, or is it only for enterprise shops?

Small agencies have an easier time selling this than large ones. The conversation is simpler: 'You should be showing up when people ask AI to recommend someone like you. You're not. Here's your score. Here's how we fix it.' That pitch works in a 20-minute discovery call. Enterprise shops get stuck in committee. The solo consultant with 10 clients can move faster and prove results before the big agencies finish their proposal templates.