Manual AI Visibility Check: Who Do ChatGPT and Perplexity Recommend Instead of You?
The AI visibility tools charge hundreds of dollars a month to show you whether ChatGPT, Perplexity and the other LLMs mention you. Before I built Mavek I did this by hand every Monday. This page has the exact steps, the Google Sheet template, the message to send the writers, and the Claude Routine that does the Monday run for you.
The Steps
- Open ChatGPT and click the temporary chat icon. Memory tilts a normal chat towards brands you have mentioned before. Temporary chat starts cold.
- Type your buyer prompt and press enter. Not your keyword. The question a buyer types. Mine is "what is the best AI tool for AEO".
- Click the three dots under the answer, then View sources. Open every link it shows.
- Repeat the whole thing 3 times in fresh temporary chats. The links change between runs; one run tells you nothing.
- Copy every link you opened into the Google Sheet. One row per link, with the run number. The template below does the counting.
- Read the Summary tab. The pages and sites cited across most runs are the ones deciding the answer. That is who ranks for your buyer prompt.
- Reach out to the writers of those pages and ask them to list your site. The message is further down.
Same method in Perplexity: paste the prompt, click Sources at the top of the answer, open every link, log them with Engine set to Perplexity. The sheet splits the counts by engine.
The Sheet Template
Make a copy of the template. Five tabs: Read me, Prompts (your brand and your buyer prompts), Links (one row per link per run), Summary (pages and sites by how often they were cited, split by ChatGPT and Perplexity, builds itself), and Outreach (who you contacted and what happened). Three example rows are in Links; delete them when you start.
Buyer Prompts to Start With
If you only have one, use the one in the video. If you want more, fill the brackets and paste into the Prompts tab.
1. What is the best [category] tool for [customer]?
2. Which [category] tool should a [customer] use in 2026?
3. [Competitor A] vs [Competitor B]: which is better for [customer]?
4. Is there a cheaper alternative to [Competitor A]?
5. What do [customer type] use for [job to be done]?
6. Best [category] tool for a team of [size]?
7. Best [category] tool under $[price] a month?
8. Which [category] tool works with [platform you integrate with]?
9. How do I [job to be done] without hiring an agency?
10. What is the best [category] tool for [industry]?The Message to the Writers
Every page in the Summary tab was written by someone. Find the byline or the contact page, and send one short note. Give them something before you ask: a fix on their page (a price that changed, a dead link, a tool that no longer exists) or a free account so they can test you. The ask is one sentence.
Subject: one fix for your [page title] post
Hi [name],
Your [page title] post keeps coming up when I ask ChatGPT and Perplexity "[your buyer prompt]", so I read it properly.
One thing has changed since you published: [the fix, e.g. "Profound no longer has the $99 Starter plan, the pricing page now shows Trial and Enterprise only"]. Source: [URL, date].
I run [your product], [one line on what it is]. If a fix like this earns a line in the post, happy to send a free account so you can test it before you write anything. If not, the fix is yours either way.
[your name]Track every send in the Outreach tab. Most writers update a list within a week if the fix is real and the ask is small.
The Claude Routine
This is the exact setup. In the Claude desktop app, turn on Claude in Chrome and the Google Sheets connector, open Routines, create one for Monday morning, and paste the prompt below. It runs the steps above in your logged-in ChatGPT and Perplexity, one tab, paced so you do not get rate-limited, writes every link into your copy of the sheet, and messages you with what changed since last Monday.
You are my manual AI visibility check. Run this every Monday at 9am.
Setup you can rely on:
- Claude in Chrome is connected. Use ONE tab only and wait at least 10 seconds between chats so ChatGPT and Perplexity do not rate-limit my account.
- The Google Sheets connector is connected. The sheet is my copy of "Manual AI Visibility Check". Tabs: Prompts (my brand in B1, buyer prompts in column A from row 3), Links (Date, Engine, Prompt, Run #, URL, Site, Mentions my brand?, Notes), Summary, Outreach.
Do this, in order, for every prompt on the Prompts tab:
1. ChatGPT: open chatgpt.com, click the temporary chat icon and confirm the "Temporary chat" label is showing. Never use a normal chat.
2. Paste the prompt exactly as written. Wait for the full answer.
3. Click the three dots under the answer, then View sources. For every link shown, append one row to the Links tab: today's date, "ChatGPT", the prompt, the run number, the URL. Leave Site blank (it fills itself). In "Mentions my brand?" write yes if the answer named my brand (Prompts!B1), otherwise no.
4. Close the chat. New temporary chat. Same prompt. Repeat until the prompt has 3 runs.
5. Perplexity: open perplexity.ai, paste the same prompt, click Sources at the top of the answer, and log every link the same way with Engine = "Perplexity". 3 runs.
6. Move to the next prompt.
When every prompt is done:
7. Read the Summary tab (it rebuilds itself). Compare with last Monday's rows in Links.
8. Send me a short message: the top 5 pages by times cited, any page cited for the first time, any page that dropped out, and whether my brand was named in any run. Nothing else.
Rules:
- Record the URLs exactly as shown. Do not shorten or clean them.
- If an engine refuses, errors or shows no sources, log one row with "no sources" in Notes and move on. Retry at most once per run.
- If Chrome loses a login, stop and tell me. Do not try to log in.
- Never type anything into ChatGPT or Perplexity other than the prompts on the Prompts tab.Two things to check the first time it runs: that the temporary chat label is showing in the screenshots it takes, and that the Links tab has 3 runs per prompt per engine. After that it is a Monday message you read with coffee.
What the Counts Tell You
| Signal | What it means |
|---|---|
| Page cited in 3 of 3 runs | That page is deciding the answer for your prompt. The writer is your first email. |
| Page cited in 1 of 3 runs | Noise. It got pulled once. Do not build a strategy around it, and do not panic if it praises a competitor. |
| Your brand named in 0 runs | You are invisible for your own buyer prompt. The Summary tab is the fix list. |
| Same site cited for several prompts | That site is a gatekeeper. Getting listed on it moves several prompts at once. |
| Sources are all vendor listicles | The category is being decided by whoever writes the comparison posts. Write one, or get into theirs. |
Where the Paid Tools Start
Profound, Peec, Scrunch and the rest run the same loop at scale: hundreds of prompts, several engines, every day, with a trend line. For a handful of prompts on ChatGPT and Perplexity, the sheet above is the same data. The gap none of the trackers close is the next step: turning the cited-pages column into changes on your site and pitches to the pages that keep showing up. That is the part Mavek does, across ChatGPT, Claude, Gemini, Perplexity and AI Overviews, and it is the part I do by hand for clients under Found and Cited.
Frequently Asked Questions
Why run each prompt 3 times instead of once?
ChatGPT's answers and sources vary between runs. A single screenshot tells you what one run pulled, not what it relies on. Three fresh runs per prompt is enough to separate the pages that come up every time from the ones that got pulled once. If you want tighter counts, run five.
Why a Temporary chat?
A normal chat uses memory from your earlier conversations and tilts the answer towards brands you have mentioned before, including your own. Temporary chat starts cold every time, which is closer to what a stranger asking the same question would see.
Do I need ChatGPT Plus for this?
No. Temporary chat and Sources are available on the free tier. The paid AI visibility tools run the same prompts through the same interface or through scraped data; the difference is volume, history and the dashboard.
What do Profound, Peec and the other tools add on top of this?
Scale and history. They run hundreds of prompts across several engines daily and keep the trend. For a handful of prompts on two engines, the sheet gives you the same answer. Where a tool earns its price is when you need 200 prompts across ChatGPT, Claude, Gemini, Perplexity and AI Overviews every week, and when you want the gaps turned into a to-do list on your site.
Can the Routine run on its own?
Yes, as a scheduled Routine in the Claude desktop app with Claude in Chrome and the Google Sheets connector enabled. It drives your logged-in ChatGPT and Perplexity in one tab, paces itself to avoid rate limits, and writes to your copy of the sheet. If you do not want to give it the browser, do the 3 runs by hand on Monday; it is about 10 minutes per prompt.
How often should I run it?
Weekly is enough. Answers do not move day to day. Weekly gives you a trend line after a month and catches a new competitor entering the answers within a week of it happening.
What should I do with the cited pages?
Treat them as the target list. A comparison post cited across several prompts is worth more than most backlinks. Read the page, find the writer, send the message above with a real fix in it, and track it on the Outreach tab. Mavek's Found and Cited work starts from exactly this list.