Why Marketing Agencies need this
Client acquisition tops the list of agency challenges, and one survey put it at 37 percent of agencies. Yet plenty of agencies leave their own site half-finished, and prospects notice when the people selling SEO have not looked after their own properties. Now a second question lands in every pitch and every monthly call: are we visible in ChatGPT? Manual rank reports already eat hours per client, and clients worry over normal fluctuations, so nobody wants another report that is hard to defend. The AI Visibility Checker gives you a narrow, defensible answer. Enter a URL and an optional brand name, and it reads up to four pages as raw HTML, plus robots.txt and llms.txt, then returns a score out of 100 with fixes. It never asks any assistant anything, so it cannot say whether a brand is mentioned, cited or recommended. It tells you whether the site is readable and well marked up for AI systems. For mentions, ask the assistants the same prompts by hand and write the answers down. Use the score for your own site first, then for clients.
Marketing Agencies specific guidance
Check your own site first
Run your agency's homepage before any client's, since a prospect who asks about AI search will look at it. Enter https://youragency.example/ and add the agency name as the brand. Open the AI Crawler Access category first. If robots.txt blocks GPTBot, PerplexityBot, ClaudeBot, Google-Extended or CCBot, the rest of the report matters less. In the next ten minutes, open youragency.example/robots.txt in a browser and look for those names, then view the page source and confirm your service copy is in the HTML. A clean robots.txt does not prove any assistant uses your pages, and the brand checks read only your own site, never third-party sites.
Reading the score for a client
The AI Visibility Score is a SeoMods criterion, not a platform rating, so label it that way in a client deck. Read the four category scores before the total: AI Crawler Access, Brand & Entity Strength, Machine-Readable Content and Trust & Freshness. The Google AI Overviews, ChatGPT Search and Perplexity bars are averages of those categories, estimated from weighted signals, not a live query. A bar at 70 is not a 70 percent chance of being named. Because the tool keeps no history, save each run's score, cards and issue list with the date, and compare a client only with their own earlier run, never with another client.
Fix order and what it proves
Work the critical blockers first: allow the AI crawlers in robots.txt, get the main content into the initial HTML, and serve the whole site over HTTPS. Next add Organization and WebSite JSON-LD, sameAs links to the client's social profiles and a linked About page. Then add named authors, publish or update dates and at least two outbound citations. Re-run after each batch so you can tell the client which change moved which category. A higher score shows the site is easier for AI systems to read. It does not show one mention in ChatGPT, so keep a separate log of prompts, dates and the brands each assistant named.
Example scenarios
1. Your own agency homepage
- Situation
- Illustrative example: a developer added a blanket bot-blocking block to robots.txt last year to stop scrapers, and it also names GPTBot and CCBot. Nobody on the sales side knows, and prospects keep asking about AI search.
- Input
https://youragency.example/with the brand name entered as the agency name- Reading the result
- You would typically see the AI bots blocked card at 2 and AI Crawler Access pulling the total down, perhaps to a score of 58 out of 100, with the robots.txt fix listed as a critical blocker. Remove the Disallow lines for the AI bots you want to allow, then re-run. Passing afterwards means the crawlers are let in, not that any assistant names you.
2. A client store built as a JavaScript app
- Situation
- Illustrative example: a client moved to a headless front end, and the homepage now loads as an empty shell of about 40 words until scripts run. The rendered page looks fine in a browser, so nobody noticed.
- Input
https://client-store.example/with no brand name- Reading the result
- Because pages are read as raw HTML, the content-in-raw-HTML check would typically fail against the 200-word threshold, and Machine-Readable Content would also look weak. The fix is server-side rendering or pre-rendering, which is a development ticket. The result does not say how Google indexes the page, only what a non-JavaScript reader sees.
3. Pitch opener for a prospect
- Situation
- Illustrative example: a prospect's homepage title says only Welcome, the brand name appears once in the body, and there is no Organization schema or sameAs link. The owner has asked whether competitors show up in ChatGPT.
- Input
https://prospect-clinic.example/with the brand name entered- Reading the result
- The brand checks would typically flag the missing brand in the title and fewer than 3 mentions on the homepage, and Brand & Entity Strength would score low. Show those rows as the fixable gap, for example a score of 41 now against a target after fixes. Be clear you have not measured any ChatGPT answers.
Common mistakes
- Calling the score a visibility figure. The number measures how readable and well marked up the site is for AI systems, not how often a brand appears in answers. Selling it as share of voice will not survive the first client question.
- Reporting the platform bars as probabilities. The Google AI Overviews, ChatGPT Search and Perplexity bars are averages of category scores, estimated rather than queried live. Present them as readiness only.
- Skipping the hand-run prompts. If the client's real question is whether ChatGPT mentions them, you still have to type the prompts into each assistant and record the answers. The checker cannot do that for you.
- Ignoring your own agency site. Pitching AI-search work while your own robots.txt blocks GPTBot costs credibility fast. Run the checker on your homepage before you run it on a prospect's.
- Treating one run as a trend. The tool stores no history, so a single score means little by itself. Save the score and issue list with the date after every run, and compare a client only with their own earlier run.
Frequently asked questions
Can the AI Visibility Checker tell me if ChatGPT mentions my client?
No. It never asks ChatGPT, Gemini, Perplexity or Google AI Overviews anything, so it cannot show whether a brand is mentioned, cited, recommended or ranked, and it cannot compare a client with a competitor. It measures whether the site is readable and well marked up for AI systems. To answer the client's question, type the same prompts into each assistant yourself, note the date, and write down which brands appear and which sources are linked.
How do I explain the platform readiness bars in a client report?
Describe them as estimates. The bars for Google AI Overviews, ChatGPT Search and Perplexity are averages of the category scores, and the tool labels them as estimated from weighted signals, not a live query of each platform. Say they show how ready the site is to be read, not how often it appears. Avoid words like share of voice or probability of citation, because the number supports neither. Put the date of the run next to the figure so next month's comparison is fair.
Is it worth using on my own agency site to get SEO clients?
Yes, as a quick self-check before pitches. Prospects often look at an agency's own site, and a poorly optimized one hurts credibility. Run your main service URL with your agency name as the brand. Confirm that AI crawlers are allowed, your copy sits in the raw HTML, and Organization schema and an About page link exist. It will not win you clients by itself, and a good score says nothing about rankings or inquiries; it only removes avoidable technical gaps from your own site.
Does it work for SEO reporting for agencies with monthly client reports?
It can supply one line in a monthly report, but it is not a reporting platform. It keeps no history, does no tracking and offers no white-label export, so save the score, category scores and issue list from each run yourself and note the date. Use it to show that specific fixes, such as unblocking GPTBot, were completed. It does not track rankings, so keep your rank tracking separate and explain normal fluctuations there, not here.
How many pages does it check, and does it run JavaScript?
It checks up to four pages in total: the URL you enter plus a small sample of the same site's pages, read as raw HTML with no JavaScript. It also reads robots.txt and llms.txt. That matters for sites built on JavaScript frameworks, because content that appears only after scripts run counts as missing. Four pages is a sample, not a crawl, so for a full technical review of a client's site use a site crawl and treat this score as one angle.
Reference: AgencyAnalytics covers client acquisition in its guide to 12 proven strategies for getting agency clients, including the point that agencies often neglect their own properties and hurt their credibility.