Why clients keep asking about ChatGPT and Perplexity

An AI visibility check shows whether a client brand appears in answers from ChatGPT, Perplexity and Google AI Overviews. Run one before the next client call, because the question is coming. B2B SaaS buyers now research in AI tools first and visit websites later.

Google rankings no longer predict AI citations. The overlap between the Google top 10 and AI citations fell from 75% to 17-38%. So a client can sit on page one and still be absent from the answer. Also, 80%+ of AI answer queries end without a click. As a result, a missing citation means the buyer never sees the brand.

Clients notice this before agencies do. They ask why a competitor shows up in ChatGPT while their own brand does not. Most agencies have no repeatable answer, so each reply turns into a one-off investigation.

What an AI visibility check should tell you

An AI visibility check should return four answers per client. Without them, you cannot explain why ChatGPT or Perplexity ignores their content.

Take a B2B SaaS client that sells invoice automation software. Run a set of buyer queries, such as "best invoice automation for mid-size finance teams," and record the following:

• Whether the brand is cited in the answer at all.
• Which pages get cited, for example the pricing page, a comparison post or nothing.
• Which competitors appear instead, and which of their pages the engine quotes.
• Which queries trigger a Google AI Overview, because those need different content than queries without one.

The result is a gap list. If a competitor's comparison post gets cited and your client has none, you know what to write next.

Why Google rankings no longer predict AI citations

Ranking is still the foundation. If Google can't rank you, AI won't cite you. But a top 10 position no longer guarantees a citation. The overlap between Google's top 10 and AI citations fell from 75% to 17-38%. Learn more about how AI Overviews work and citation mechanics.

So a page can hold position three and still be missing from ChatGPT answers. Also, 80% of AI answer queries end without a click. A missed citation therefore leaves no trace in your traffic reports.

For agencies, this changes what a client report needs to show. A rankings-only report tells the client where they sit on Google. Then the client asks why a competitor appears in Perplexity, and the report has no answer. An AI visibility check, run next to rank tracking, gives you that answer before the client asks.

7 steps for an AI visibility check on a client site

An AI visibility check takes about 60 minutes per client when you follow a fixed routine. Run it once a month, and you have a clear answer when a client asks why ChatGPT never mentions them. Keep one spreadsheet per client so results stay comparable from month to month. Each step below says what to do and what to record, and gives an example. The example client is a fictional invoicing tool for marketing agencies. Because the routine never changes, a junior team member can run it after one walkthrough.

1. Write the prompt list (8 minutes)

Pull 15 to 20 questions that buyers ask before they choose a tool. Good sources are Search Console queries, sales call notes and support tickets. Mix category searches, comparison questions and problem-based questions. Record each prompt in the sheet with a funnel tag: TOFU, MOFU or BOFU.

Then freeze the list. If you change prompts every month, you cannot compare results. For the invoicing client, the list includes "best invoicing software for marketing agencies" and "invoicing tool with retainer billing". It also includes "[client] vs [competitor]". Add two or three new prompts per quarter and mark them as new, so the trend line stays honest.

2. Run every prompt in three engines (10 minutes)

Open ChatGPT, Perplexity and Google, and paste each prompt. Use a fresh or logged-out session, because personal history changes the answers. Record the engine, the date and one of three outcomes: cited with a link, mentioned without a link, or absent.

Answers vary between runs, so run your five most important prompts twice. Then log the better and the worse result. As an illustration, the invoicing client might be named by ChatGPT on 2 of 18 prompts and cited by Perplexity on 4. Google might show an AI Overview on 9 of 18 queries. These numbers become the baseline you report against next month.

3. Record which competitors appear (8 minutes)

For every prompt, write down each brand the engine names, in the order it appears. Add a column for your client's position, or "none". Competitors that show up repeatedly are the ones buyers see first.

Also note how the engine describes each brand. A competitor tagged "best for agencies" owns a positioning your client may lack. In the invoicing example, two rivals appear on 12 of 18 prompts, and both are described as "built for agencies". The client's own site says "for growing teams". That mismatch gives you a concrete fix for the content plan.

4. List the sources the engines cite (8 minutes)

Perplexity and Google AI Overviews show their sources. ChatGPT does too when browsing is active. Copy every cited URL into a second tab and tag it by type: competitor page, review site, forum thread, news or the client's own site.

The tags show where AI engines look for answers in this category. If review sites dominate, listing profiles matter as much as blog posts. For the invoicing client, imagine that 7 of the 11 cited sources are comparison articles on third-party blogs, and only one is a competitor's own page. That tells you to publish comparison content on the client's site and pursue mentions elsewhere.

5. Compare with Google rankings (10 minutes)

Open Search Console and pull the position for each prompt's closest keyword. Record the average position and the ranking URL. Overlap between Google's top 10 and AI citations has fallen to between 17% and 38%, so a good ranking does not guarantee a citation. Still, pages that cannot rank rarely get cited at all.

Sort the results into four groups: ranks and cited, ranks but not cited, cited but not ranked, and neither. For the invoicing client, a page at position 6 for "retainer billing software" may never appear in answers. That page goes on the review list for step 6.

6. Audit the pages that should be cited (8 minutes)

Open the pages from the "ranks but not cited" group. Check whether the answer sits in the first paragraph, because engines extract it from there. Look for FAQ sections with schema markup, article schema and an author with a named bio. Also check for statistics, lists and quotes, since structured pages tend to earn 30 to 40% higher AI visibility.

Record a pass or fail per element in the sheet. The invoicing client's retainer billing page might open with 200 words of background and have no FAQ. As a result, it fails two checks. Rewriting the opening and adding five question-and-answer pairs is a half-day fix.

7. Score the month and set next actions (8 minutes)

Count the prompts where the client is cited, mentioned or absent, and calculate a citation rate. Compare it with last month. Then write three actions for the coming month, each tied to a finding from steps 3 to 6. Send the client a one-page summary with the rate, the trend and those actions.

Manual checks work well for one or two clients. At five or more, AI visibility tracking tools save time by running prompts on a schedule. Whichever AI visibility tracking tool you choose, keep the manual sheet for the first two months. It shows you what the tool's numbers actually mean.

Step 1: List the queries your client's buyers ask

Every AI visibility check starts with a fixed list of buyer questions. Build 10 to 20 per client. Fewer gives noisy results, and more eats your hours. Cover category searches ("best project management software for agencies"), comparison searches ("Tool A vs Tool B") and problem-based searches ("how to reduce churn in a SaaS trial").

Take the questions from data. Run a seed keyword through a keyword research tool, then check Google Search Console for queries the client's site already ranks for. Those queries show what buyers actually type, so they make better prompts than anything you invent.

Then lock the list. If the questions change each month, you cannot tell whether visibility moved or the prompts did. Keep the same set for every report and add new questions as a separate batch.

Step 2: Run each query in ChatGPT, Perplexity and Google

Run the same query list in all three engines during one sitting. For each result, log whether the client is named, linked or absent. Then record the date and the engine, because a result without those two fields cannot be compared later.

Answers vary between runs. ChatGPT and Perplexity generate a new response each time, so the sources they cite shift. Google AI Overviews also appear for only some queries, so a missing overview says nothing about the client.

To handle this, repeat each AI visibility check three times, ideally on different days. Then count how often the client appears. A client named in two of three runs has a stable mention, while one named once is noise.

Step 3: Record which competitors get cited

Next, write down every brand and URL the engines cite for each query. Use one spreadsheet per client, with a row per query and a column per engine. This keeps the AI visibility check repeatable, so you can rerun it next month and compare results.

Then look for patterns. For example, one competitor may appear on most comparison queries, while a review site owns the "best tools" prompts. Also note which page type gets cited: a pricing page, a blog post or a third-party listicle.

• Queries where a competitor is cited and your client is absent
• Queries where the engines cite only third-party sites

Each of these patterns points to a content gap. Because the client can fill a gap with one targeted article, you end up with a prioritised list to bring to the next client call.

Step 4: Check which client pages get cited

Open the citations from your AI visibility check and sort each one by page type. A cited URL is a blog post, a product page, or a third-party site such as a review directory or a forum thread. Each type tells you something different, so group them before you draw conclusions.

Then flag every query where no client page appears and none exists to appear. These queries are content gaps. AI engines cannot cite a page that was never published, so the fix is a new article.

As a result, each flagged query becomes a line in the content plan. Add the keyword, the funnel stage and the client, then schedule it. Third-party citations still matter, but they point to a different job: getting your client mentioned on those sites.

Step 5: Compare against Google rankings

Open Search Console and pull the same queries you tested in ChatGPT and Perplexity. Then match each query to your client's average position. This turns the AI visibility check into a diagnosis.

Sort the results into two groups:

• The client ranks on Google, but no AI engine cites it. The page exists, so the fix is structure: answer-first openings and FAQ schema.
• The client neither ranks nor gets cited. Here the page needs basic SEO work first, because AI engines rarely cite what Google ignores.

You need both views because the overlap between Google's top 10 and AI citations has dropped from 75% to 17-38%. A strong ranking no longer predicts a citation. As a result, a Google report alone hides the gap your clients are asking about.

Step 6: Audit page structure for citation

An AI visibility check should cover the pages themselves as well as prompts and mentions. AI engines pull from content they can extract quickly, so structure matters. Pages with structured lists, quotes and statistics show 30-40% higher AI visibility.

Score every client page pass or fail on these points:

• The first paragraph answers the page's main question directly.
• FAQ schema is present and matches the visible questions.
• At least one structured list appears in the body.
• A sourced statistic backs a main claim.
• A named quote or expert source appears.
• Author schema names a real person.

Any page that fails two or more points goes to the top of the fix list. Then re-check it after the update.

Step 7: Score the results and set next actions

Turn the log from your AI visibility check into one number. Divide the queries where the client is cited by the total queries tested. For example, 6 citations across 20 queries gives a score of 30%. Track it monthly so the trend shows progress.

Then assign three next actions per client, because a score without a task changes nothing. Pick from this list:

• Publish an answer-first article for a query where a competitor is cited instead.
• Add FAQ schema to a page that ranks on Google but never appears in AI answers.
• Update an older article that cites outdated figures.

Finally, cap the routine at 30 minutes per client each month. Because the time is fixed, the check stays inside your margin. If a client needs more, bill it as a separate task.

What to look for in AI visibility tracking tools

A manual AI visibility check works for two or three clients. You type prompts into ChatGPT and Perplexity, then log the answers. Past five clients, that routine eats hours you cannot bill.

So judge AI visibility tracking tools on the criteria that matter at agency scale:

• Separate workspaces for each client domain.
• Pricing that does not climb with every extra client or seat.
• Coverage of ChatGPT, Perplexity and Google AI Overviews.
• Prompt tracking over time, because answers change between runs.
• Exports you can paste straight into a client report.

Also check vendor claims against third-party reviews on G2 or Trustpilot. Finally, ask what the tool does after it finds a gap, because a report alone fixes nothing.

Pricing that holds up across many client accounts

Per-seat and per-project pricing grows with every client you sign. A tool that charges EUR 100 per project costs EUR 500 for five clients and EUR 1,000 for ten. But your own fee per client stays the same, so margin shrinks with each new account.

Multi-domain plans change the math. RankRealizer's Scale plan costs EUR 200 per month and covers up to 5 domains. The cost per client works out like this:

• 5 clients: one Scale plan at EUR 200, so EUR 40 per client.
• 10 clients: two Scale plans at EUR 400, so EUR 40 per client.

Because the cost per client stays flat, a new account adds a known amount instead of a new seat fee. Each client also gets a separate Knowledge Base, so brand voice and forbidden words do not blur across accounts. As a result, you can price retainers with confidence.

Prompt coverage, engine coverage and repeat runs

Before you pay for a tool, check four things. A one-off AI visibility check tells you little if the setup behind it is thin.

• Which engines it queries. ChatGPT, Perplexity and Google AI Overviews are the minimum, because your clients ask about all of them.
• How many tracked queries each client gets, and whether the cap is per client or shared across your account.
• How often checks run: daily, weekly or only on demand.
• Whether the tool stores history, so you can show a client a trend line.

Repeat runs matter because AI answers vary. The same prompt can name different brands on Monday and Thursday. So a single result tells you very little.

Instead, measure how often your client appears across many runs. That share is a number you can report, and it holds up when a client questions it.

Reporting that clients can read

A client needs a few plain numbers: how often their brand appears in AI answers, which queries trigger it, and whether that changed since last month. Raw logs only create more questions. So check what a tool exports before you commit.

Look for PDF or CSV exports, white-label options that carry your agency branding, and clear before and after views. A side-by-side comparison of this month and last month answers most client questions at a glance. Without it, you rebuild the story in a spreadsheet every time.

Also time the process. If a report takes 45 minutes per client, ten clients cost you nearly eight hours a month. Good tooling cuts that to a few minutes, because those hours come straight out of your margin.

How to choose one AI visibility tracking tool for all your clients

Begin with the decision you need to make, because a feature list will not make it for you. Ask which tool lets five clients share one workflow without a new seat fee for each. Then test it.

Run the same five clients through two options for two weeks. Use identical dates and identical prompts. Record three things:

1. Hours spent per client on setup and reporting.
2. Total cost, including extra seats and projects.
3. Clarity of output, meaning whether you can explain it to a client in five minutes.

A report showing ChatGPT ignoring a client still leaves you to write the content that earns a citation, so tracking alone does not fix visibility. Favour the option that connects to content production, because that is where your hours and margin go.

Run a two-week trial on three client accounts

Pick one small client, one mid-size client and one difficult client. The difficult one might sit in a crowded category or have thin content. Because it exposes weak spots, it tells you more than the easy accounts.

Before you start, write down the rules:

• List 10 to 15 tracked queries per client, such as questions buyers ask ChatGPT or Perplexity.
• Run the same AI visibility check on day one and again on day 14.
• Log every hour spent per client, including editing.
• Set pass criteria first, for example citations gained and editing time per article.

Then test without commitment. RankRealizer offers 3 free articles with no credit card, and pay-per-use costs EUR 5 per article. So the whole trial stays cheap, and you decide with your own numbers.

Decide on tracking only or tracking plus content

An AI visibility check shows where a client is missing from ChatGPT or Perplexity answers. A pure tracker stops there. It lists the gaps, but the agency still writes every fix.

That is a problem for a small team, because the editing hours come straight back. Generic drafts need heavy rework, so each gap costs time you already lack.

Tools that also produce structured content close the loop. RankRealizer, for example, generates articles with answer-first structure and FAQ schema, then publishes to WordPress or Webflow. Still, you give up some control over wording, since the output works best when you review and approve it.

Tracking only makes sense if each client already has a writer. If not, pick a tool that also writes the fix.

Fixing what the check finds: structure content for citation

An AI visibility check means nothing if you don't act on it. Most AI-generated content has no SEO structure, so it doesn't rank or get cited. Agencies end up rewriting it from scratch. Structure it right from the start, and the edits become smaller.

• Answer the question in the first paragraph. AI engines pull from the top of the page.
• Build in a schema-ready FAQ with short question and answer pairs.
• Add Article and Person/Author schema so E-E-A-T signals are readable by machines.
• Use lists with actual numbers. Pages with structured lists, quotes and statistics get 30-40% higher AI visibility.
• Cover the keyword gaps your clients' competitors already filled.

RankRealizer adds these elements to every article automatically. Competitor intelligence shows you the gaps, and one click builds a response article. Your team reviews and approves rather than rewriting from the ground up.

Where RankRealizer fits in an agency workflow

RankRealizer is a GEO tool built with RhineWeb and SaphirSolution GmbH, two SEO agencies. So its output is structured for Google ranking and AI citation. It is not a technical SEO audit tool. Still, it works best when you review and approve articles instead of rewriting them.

• Each client gets a Knowledge Base with brand voice, ICP and forbidden words.
• Every article runs through 9 steps, including SEO scoring and an originality check.
• Then articles publish automatically to WordPress or Webflow.
• Also, competitor monitoring runs per domain.
• Finally, the dashboard shows AI impressions and AI clicks.

Pricing stays predictable because you pay per plan or per article. Scale costs EUR 200/month for up to 5 domains, and Growth costs EUR 100/month. Pay-per-use is EUR 5 per article. The free trial gives you 3 articles, with no credit card.

Reporting AI visibility to clients every month

Run the monthly AI visibility check in 45 minutes or less per client. Use the same query list every time. Changing queries breaks your comparison month to month. Then update the citation share and note which new articles went live.

Keep the report to one page with four parts:

1. Citation share this month versus last month
2. New articles published and their target queries
3. Queries where competitors are cited and your client is not
4. Next steps for the coming month

When a client asks why they are not showing up in ChatGPT, show them the report. The AI visibility check surfaces the exact queries where competitors appear and the client does not. That gives you a real answer instead of a guess. Then point to your content plan. Each gap becomes an article with a publish date. Next month's report shows whether the citation share moved.

Run your first AI visibility check this week

You can finish a first AI visibility check in about an hour, so pick one client and start there. Choose a client who has already asked why they are missing from ChatGPT or Perplexity.

1. Write 10 queries a buyer would type, such as "best project management software for agencies".
2. Run each query in ChatGPT, Perplexity and Google AI Overviews.
3. Log whether the client is cited, which competitors appear and which pages get named.

Then decide how to track it. A spreadsheet works for one or two clients, but it gets slow across ten. If that happens, compare each AI visibility tracking tool on price per project before you commit.

Also, RankRealizer offers 3 free articles with no credit card. You can use them to test content production for that same client at no cost.

Frequently Asked Questions

What is an AI visibility check?

An AI visibility check tests whether ChatGPT, Perplexity and Google AI Overviews mention or cite a brand when buyers ask category questions. You run the questions your client's buyers would ask, then record which brands and sources appear. The result shows where the client is cited, where competitors show up instead, and which pages need work.

How do I run an AI visibility check for a client?

Start with 10 to 20 questions a buyer would ask about the client's product category. Then enter each one in ChatGPT, Perplexity and Google, and note whether the client is named or linked. Because answers change between runs, repeat each question a few times and log the results in one sheet per client.

Why can a client rank on Google and still fail an AI visibility check?

The overlap between Google's top 10 and AI citations has dropped from 75% to 17-38%. As a result, a page can rank well and still be missing from AI answers. Structure also matters, because pages with lists, quotes and statistics show 30-40% higher AI visibility. Answer-first paragraphs and FAQ schema help AI engines extract a clear answer.

What do AI visibility tracking tools do?

AI visibility tracking tools run a fixed set of prompts on a schedule and log mentions, citations and competitor appearances. They replace manual spot checks, so you can show clients a trend instead of a screenshot. Still, check the price per project before you commit, because the cost multiplies with every client you add.

How do I choose an AI visibility tracking tool for a small agency?

Look for multi-client support, a flat price per domain and reporting you can send to clients without rework. Tracking only shows the gap, so you also need a way to close it. RankRealizer's Scale plan covers up to 5 domains for EUR 200 per month. It also gives each client a separate Knowledge Base and a dashboard with AI impressions and AI clicks.

How often should I run an AI visibility check?

Run one monthly for each client, and again after you publish a batch of new articles. AI answers vary between sessions, so repeat each prompt at least twice before you log a result. A consistent schedule also makes month-over-month changes easy to report.