How to sell AI visibility services

Your clients and the owners you talk to have started asking why ChatGPT names their competitors and not them, and you would like to sell that work. Here is how to sell AI visibility services when it is just you: a paid check that shows a business how often AI answers name it, a monthly add-on priced from your hours, and 100 personal messages a day that open with the owner's own result. Most agencies either promise a place in ChatGPT they cannot control, or never offer the work at all.

Do these three things

About 5 hours a day, every day, next to client work. Judge your messages after 2,800, and each client's count at week 12, never on one answer.

Show the three actions in fullHide the full list

1. 1Today: write a one-page AI visibility offer: a paid check of 20 buyer questions in ChatGPT, Perplexity and Google AI Mode, a monthly add-on priced from its hours, and the line "We never promise a mention".
2. 2Every day: send 100 personal messages to owners in your niche that open with what ChatGPT said when you asked their buyer's question. AI drafts; you check the answer yourself, and offer to show the full check on a 20-minute call.
3. 3Every Sunday: ask each client's 20 questions in the 3 tools, put "named in X of 60" in their monthly report, and count your messages, calls and checks sold per 100 messages.

Get help doing these

ChatGPT or Claude reads this guide and works through it with you.

Do this with ChatGPT(opens in a new tab)Do this with Claude(opens in a new tab)

1. Write your AI visibility offer on one page

AI visibility services help a business get named more often when its buyers ask ChatGPT, Perplexity or Google's AI answers for a recommendation. Write your offer on one page today: the check you run first, the work you do each month, the number you report, and what you will never promise.

Use the owner's words, not yours. Owners do not search for "GEO" or "AEO", and they have rarely looked at AI Overviews on purpose. They know "when people ask ChatGPT for the best [what they sell], are we named?". Put that sentence at the top.

The check. Write 20 questions a buyer would ask before they find the business: 18 about the problem or the kind of product, with no brand name, and 2 about the business itself. This is the same list as in tracking AI mentions. Ask each one in ChatGPT, Perplexity and Google AI Mode, 3 times each, in a fresh chat. That is 180 answers. Log whether the business was named, who was named instead, and which sites each answer cited.

Ask many times because one answer proves nothing. In SparkToro's January 2026 study of 2,961 answers, ChatGPT and Google's AI gave the same list of brands less than once in 100 tries. The study said the share of answers that name a brand, over many tries, is a reasonable number to track, and any "position in AI" is not.

The monthly work. It is the work in getting a brand mentioned by ChatGPT: pages on the client's site that answer buyer questions, answers in the threads AI cites, reviews on the sites AI reads, and pitches to the lists AI tools cite. For a client who sells to businesses, use the steps for getting recommended as a supplier. Add the AI Overviews work if the client's buyers search on Google first.

What you promise. Promise the work, the count and the report. Never a mention, a position or a date. Nobody controls what an AI tool says. Since February 2026, ChatGPT also shows ads to some free users in the US, and OpenAI says "Ads do not influence the answers ChatGPT gives you", so a client cannot buy their way into the answer either.

Leave llms.txt and schema markup out of the headline. Google says you do not need special files or markup to appear in AI Overviews or AI Mode. It is one hour of setup inside the work, not the product.

When [their buyers] ask ChatGPT for the best [what they sell], are you named?The check (once): I write 20 questions your buyers ask, with you. I ask each one in ChatGPT, Perplexity and Google AI Mode, 3 times, and count how often you are named, who is named instead, and which websites the answers use. You get the count, the sheet, the 5 sites that matter most and a 5-minute video. $[price], ready in [5] working days.Each month: [number] pages that answer your buyers' questions, answers in the [number] threads AI uses most, and pitches to the [number] lists AI cites. You approve everything before it goes out.Every Sunday: I ask the same 20 questions again and send you one line: named in [X] of 60 answers, and the sites the answers used.What I never promise: that ChatGPT or any AI tool will name you, at what position, or by what date. Nobody controls that. I promise the work, the count and an honest report.$[price] a month, added to your current package. Month to month after the first 3 months.

• Tool: Google Docs.
• Time: 1 hour.
• You will have: one page you can paste into a message, a proposal or your site.
• It worked if: someone outside marketing can say in one sentence what they get each month, and no line promises a mention, a position or a date.
• Common mistake: selling a file ("we install llms.txt") or a promise ("we'll get you into ChatGPT"). The first does little on its own. The second you cannot keep, and it ends the client when the answers change.

2. Price the AI visibility audit and the monthly add-on from your hours

Price the AI visibility audit (the check) and the monthly add-on the way you price everything else: the hours each takes × your lowest price per hour ÷ 0.65, rounded up to the next $100. Time each part of the work before you write a price on the page.

Your lowest price per hour comes from pricing your digital marketing services. Then time the work:

• The check: about 1 hour to agree the 20 questions, about 3 hours to ask and log 180 answers at a minute each, 1 hour to sort the cited sites and pick the top 5, and 1 hour to write it up and record the video. About 6 hours.
• The monthly add-on: 4 Sunday counts of 60 answers (about 1 hour each), plus the pages, answers and pitches you put on the page, each timed. Start small, about 10 hours a month, and add hours only when a client asks for more.

Do not copy the price ranges in vendor guides. In 2026, Similarweb suggested $500 to $5,000 for an audit and $1,500 to $5,000 a month for tracking, and Superlines $1,500 to $5,000 for an audit. Both sell tracking tools, and their ranges are for bigger companies than your clients. A range does not know your hours.

Sell it as an add-on to your monthly package, not as a second package, so a client can raise what they pay you one step at a time, and write it into the client's scope of work with its numbers. If you buy in SEO work from someone else, the same sum applies to white-label SEO.

• Tool: a Google Sheet.
• Time: 30 minutes.
• You will have: two prices on your one page, each with its hours.
• It worked if: each price is at least hours × your lowest price per hour ÷ 0.65, rounded up to the next $100, and you can say the hours behind every line.
• Common mistake: a free check for everyone. A free check is for your current clients in step 3, and a short free video is for prospects. The full check is 6 hours of work, so new buyers pay for it.

3. Run the check on your current clients first and offer them the add-on

Your current clients are the easiest first buyers, because they already trust you. This week, run the full check on each one for free, send them the result, and offer the monthly add-on.

Do it with the client, as part of your normal client onboarding if they are new. Send them the 18 + 2 question template and ask them to change any question their buyers would not ask. Then ask each one 3 times in each tool, and log every answer the same way.

For a first look at who AI cites in their market, put their biggest competitor's name into Ahrefs' free AI visibility checker. It needs no signup, and it shows the top 5 domains and pages that ChatGPT, Gemini, Perplexity, Copilot and Google's AI cite next to that brand. Write those 5 down next to your own count.

Then send the result in one message, with a 5-minute Loom video (the free plan allows 25 videos of up to 5 minutes). Say the count, the names AI gave instead, the 3 sites that came up most, and one thing you would do first. Offer the add-on with its price and a start date, and say that every page and pitch goes through your usual client approvals first. Keep each client's questions and wording in their own sheet, the same way you keep client brand voices separate.

If your clients are local businesses: write the questions the way people ask for a local service: "Who is the best [service] in [town]?". In BrightLocal's 2026 survey, 45% of US consumers used ChatGPT or other AI tools for local recommendations, up from 6% a year before. A good Google Business Profile and Google reviews are part of the monthly work.

• Tool: ChatGPT, Perplexity and Google AI Mode (all free), the Ahrefs free checker, a Google Sheet, and Loom (free).
• Time: about 6 hours a client, once.
• You will have: a starting row for each client, "named in X of 180" with the date, and an offer sent.
• It worked if: every client has the dated row and has answered your offer within 7 days. A yes is good; a "not now" with a month is fine.
• Common mistake: sending a screenshot of one answer. The next time the client asks, the answer is different, and your proof is gone.

4. Send 100 messages a day that open with a prospect's own AI answer

Every day, send 100 personal messages to owners who fit your ideal customer profile. Each one opens with what ChatGPT said when you asked their buyer's question, and offers a free 5-minute video or the full check on a 20-minute call.

One question checks a whole category. Ask ChatGPT "What is the best [kind of software] for [buyer]?" once, in a fresh chat, and every business in that category is in or out. Write the date and the exact question in your lead list, then ask it twice more and in Perplexity, so you only write to owners who were left out every time. 100 a day needs about 3,000 owners a month, so add 100 new rows every day and find each owner's email.

Let AI draft, you check. Paste 10 rows into ChatGPT or Claude, each with the owner, the business, the question and the names the answer gave, plus the template below. Ask for one message per row. Then open each answer again yourself before you send. A message that gets the answer wrong ends the conversation.

Say the fact, then stop. "You were not named" is a fact the owner can check. "You are invisible to AI" is a scare, and owners delete scares. If you checked many businesses, add the count: it is your social proof, as in getting clients without case studies. A short testimonial from a client you already run the check for is even better.

Spread the 100, never cut it. Send cold email from a warmed-up second domain, never from the address your clients use, and send the rest as LinkedIn messages to people you are connected to, or as phone calls. Getting agency clients on LinkedIn covers that side, and it is the same daily target as getting your first agency client. The agency cold email guide has the limits for each. In the US, the CAN-SPAM Act "makes no exception for business-to-business email": every email needs your postal address and a way to say no, and you must honour a no within 10 business days. Each email that breaks the rules can cost up to $53,088. Follow up with everyone who does not answer, 3 times, on day 3, 7 and 14.

Subject: [their company] in ChatGPTHi [first name],I asked ChatGPT "[the exact question]" 3 times on [date]. It named [number] [tools / businesses], including [one competitor]. [Their company] was not one of them.[One honest line of proof: "I checked [number] [kind of business] the same way, and [X] were not named." or "I run this check for [number] clients every Sunday."]Want a free 5-minute video showing the 2 websites the answers used most? No call needed.[Your name], [agency]
[Postal address]
Not for you? Reply "no" and I won't write again.

• Tool: ChatGPT and Perplexity (free), your sheet, Gmail from warmed mailboxes, LinkedIn and your phone.
• Time: about 4 hours a day: about 2 hours to check and add 100 rows, 1 hour 40 minutes to send, and the rest for follow-ups.
• You will have: 700 messages a week, each logged with the question, the date and the follow-up dates.
• It worked if: 100 go out every day, each with an answer you saw yourself in the last 2 days, and fewer than 2 in 100 bounce. After 2,800 messages, at least 1 in 100 has become a conversation: a yes to the video, a real question back, or a call.
• Common mistake: one answer from one chat, or a fact AI suggested that you never saw. If replies stay low, fix the list before the words. The owner asks the same question, gets a different answer, and stops trusting you.

5. Show the check on a 20-minute call and promise the work, never a mention

On the call, run 3 of the prospect's questions live while they watch, show them the sheet, and send a written price within 24 hours. Say out loud, and in writing, what you will do and what you will never promise.

Show, then ask. Share your screen, open a fresh chat, and ask 3 of their buyer questions. The owner sees the answer change, which is the best reason to count 180 answers instead of one. Then ask the discovery questions: who their buyers are, where their last 5 customers came from, and what one more customer a month is worth.

Then offer the check at its price, and the add-on after it. Log every call and its next step in your sheet, or in a CRM once the sheet gets slow. If they already pay you for content or SEO, the add-on goes on the same invoice. Send a written proposal within 24 hours, with your one page from step 1 inside it.

Most calls bring the same three questions. Write your answers once:

"Isn't this just SEO?": Partly. Google says the basics for its AI answers are the same as for search, so I keep doing that. The new part is counting what AI says about you every week, and getting you onto the other websites AI reads, like review sites, threads and lists."Can you guarantee we'll show up in ChatGPT?": No, and nobody can. The answers change every time you ask. I promise the work, the same 20 questions every Sunday, and an honest count. If the count has not moved by week 12, I show you why and what we change."How will we know it worked?": Three numbers: how many of 60 answers name you each Sunday, how many of the sites AI uses mention you, and how many new customers say they found you through AI. You see all three in your report.

Never invent a result or a client. If you have none yet, say so, and let the live check be the proof.

• Tool: a screen share, your sheet, and Google Docs for the proposal.
• Time: 20 minutes a call, and 30 minutes for each proposal.
• You will have: a proposal with the check, the add-on and the promise lines, sent within 24 hours of every call.
• It worked if: every call gets a written price within 24 hours, and after 10 calls at least 2 have bought the check.
• Common mistake: saying "we'll get you into ChatGPT in 90 days" to win the deal. When week 12 comes and the answers have not moved, the client leaves and tells others, and losing clients costs more than the deal was worth.

6. Count client mentions and your own sales every Sunday

Every Sunday, ask each client's 20 questions in the 3 tools once (60 answers), put the count in their report, then fill in one row for your own sales. Use the rules below, judged by volume, not by one good or bad week.

Each client. Fill in one row: named in X of 60, who was named instead, and how many of the cited sites now mention the client. Add the AI visits and "found you through AI" answers from tracking AI traffic, and Google's numbers from Search Console. Put the row in their monthly report with one line on the work done. If the client asks why nothing moved in week 3, the same reasons as explaining SEO timelines apply: other websites have to change first.

Your sales. Fill in one row: messages sent, conversations, calls, checks sold and add-ons sold. Count where each sale came from: a current client, email, LinkedIn or a call. Once you have a client's numbers moving, ask whether you can write it up as a case study, and use it to market your agency.

After 2,800 messages: a source under 1 conversation per 100 → move its share of the 100 to the source with the most. Every source under 1 → change one part, in this order: the list, the first line, then the offer.After 10 calls: fewer than 2 bought the check → offer a smaller check (10 questions in 2 tools), priced from its hours.Each client at week 12: the count went up, or more cited sites mention them → keep the work. Still 0, and no cited site mentions them → move next month's hours to the 3 sites AI cites most.A client asked for more 3 times → price the extra work as a bigger add-on, from its hours.

• Tool: your Google Sheet.
• Time: about 1 hour per client and 30 minutes for your own row.
• You will have: one row per client and one sales row each week.
• It worked if: every row is filled every Sunday, and each rule has a decision written next to it when it comes up.
• Common mistake: judging a client at week 4, or your messages after 300. Answers change every run, so only weeks of counts tell you anything, and a few hundred messages are too few.

Where Scout7 helps

Scout7 helps with steps 3 and 6, the delivery work for a client who has signed.

Step 3. Scout7's AI question coverage collects the questions real buyers ask AI assistants in the client's category and shows who answers them today on Google and in AI answers. That gives you a first draft of the 20 questions, which you and the client then change. Its search and AI audit crawls the client's site and scores site health, Google Search presence, AI answer presence and answer depth, with a fix list that includes checks of robots.txt, llms.txt and structured data, and it builds a competitor set showing which rivals win the answers you want. Use these next to your own count of 180 answers, not instead of it. Scout7 diagnoses the site; you or the client make the fixes.

Step 6. Each client brand gets its own workspace, with its own brand profile, connections and calendar, switched from the top of the app, so one client's questions and audit results stay apart from another's. Run the audit again next to your Sunday counts to see which fixes are done. How many workspaces you get depends on your plan. Nothing goes live without your approval.

Your AI visibility sales checklist for the next 4 weeks

0 of 13 done

Work through this list with an AI assistant

It drafts the work and keeps track of what is done.

Do this with ChatGPT(opens in a new tab)Do this with Claude(opens in a new tab)

Step 1 · Write your AI visibility offer on one page

• Write your one-page AI visibility offer from the template.

Step 2 · Price the AI visibility audit and the monthly add-on

• Time the check and the add-on, and price both from your hours.

Step 3 · Run the check on your current clients first

• Ask each client's 20 questions in ChatGPT, Perplexity and Google AI Mode, 3 times each, and log every answer.
• Send each client their result and offer the add-on.

Step 4 · Send 100 messages a day that open with their AI answer

• Every dayCheck 100 new businesses in ChatGPT and add them to your list.
• Every daySend 100 personal messages that open with the owner's own AI answer.
• Every dayFollow up with everyone who has not answered, on day 3, 7 and 14.

Step 5 · Show the check on a 20-minute call

• Write your answers to "Isn't this just SEO?", "Can you guarantee it?" and "How will we know?".
• Every dayRun 3 questions live on every call and send a proposal within 24 hours.

Step 6 · Count client mentions and your own sales every Sunday

• Every SundayAsk each client's 20 questions in the 3 tools and put the count in their report.
• Every SundayFill in your own sales row.
• After 2,800 messagesMove the share of any source under 1 conversation per 100 to the best one.
• At week 12Decide each client's next month from 12 weeks of counts.

Copy listDownloadReset