Seoptist vs manual ChatGPT checks
Typing your own prompts into ChatGPT is where most businesses start measuring AI visibility. Here is where that works, where it misleads you, and what a tracked, sampled approach adds.
At a glance
| Criteria | Manual ChatGPT checks | Seoptist |
|---|---|---|
| Engines covered | Whichever you open, usually one | ChatGPT, Gemini, Perplexity, Claude, Google AI Overviews, AI Mode, Copilot (3 on Starter, 5 on Growth, all 7 on Agency and Enterprise) |
| Sampling | One answer per prompt, when you remember | 3 answers per prompt per engine every week (5 answers and daily runs on Enterprise; daily tracking add-on for 20 prompts at £49/month) |
| Accuracy alerts | Only if you notice a wrong price or address yourself | Answers compared against your fact card (prices, address, phone, product details) with alerts on mismatch, Growth and above |
| Audit | None | Technical SEO, Core Web Vitals and GEO readiness audit (100 pages Starter, 500 Growth and Agency, 5,000 Enterprise) |
| Checklist | Whatever you write down afterwards | One prioritised list scored impact x 2 minus effort, with how-to-fix guidance and copyable code |
| White-label | Not applicable | Agency plan (£399/month) and Enterprise: logo and colours, client workspaces, client-viewer logins |
| Price from | Free, plus your staff time | £49/month ex VAT, 14-day trial with card, two months free on annual billing |
Almost everyone starts the same way: open ChatGPT, type "best accountant in Leeds", and see whether your name comes up. It is free, it takes a minute, and it tells you something. The question is whether what it tells you is reliable enough to make decisions on, and whether you can keep doing it every week across several engines without it quietly becoming someone's part-time job.
Where manual ChatGPT checks work
For a first look, manual checks are genuinely useful. If you have never asked an assistant about your own business, a handful of prompts will show you the shape of the problem: whether you are mentioned at all, who the assistant recommends instead, and whether it gets basic facts right. That is a sensible thing to do before paying for anything, and it costs nothing but ten minutes.
Manual checks also make sense for a one-off sanity test after a specific change, such as confirming that a rebrand or a new address has reached the assistant, or for reading the full tone of an answer rather than a summary of it.
Where it breaks down
The first problem is variance. Ask the same engine the same question three times and you will often get three different lists. A single answer is one sample from a distribution, so "we were mentioned on Tuesday" says little about whether you are usually mentioned. Without repeating each prompt and recording the share of answers that include you, you are reading noise.
The second is personalisation and memory. A logged-in consumer account carries chat history, custom instructions and location signals. If you have asked ChatGPT about your own company before, it may well remember, which makes your checks look better than a stranger's would.
The third is scope. ChatGPT is one engine. Gemini, Perplexity, Claude, Google AI Overviews, AI Mode and Copilot each retrieve from different sources and recommend different businesses. Checking one tells you little about the others.
The fourth is continuity. Nobody records fifty prompts across five engines every week by hand, screenshots the answers, notes which competitors appeared, extracts the citations and compares the prices quoted against the real ones. The checks happen for two weeks and then stop.
Side by side
| Criteria | Manual ChatGPT checks | Seoptist |
|---|---|---|
| Engines covered | Whichever you open, usually one | ChatGPT, Gemini, Perplexity, Claude, Google AI Overviews, AI Mode, Copilot (3 on Starter, 5 on Growth, all 7 on Agency and Enterprise) |
| Sampling | One answer per prompt, when you remember | 3 answers per prompt per engine every week (5 answers and daily runs on Enterprise; daily tracking add-on for 20 prompts at £49/month) |
| Accuracy alerts | Only if you notice a wrong price or address yourself | Answers compared against your fact card (prices, address, phone, product details) with alerts on mismatch, Growth and above |
| Audit | None | Technical SEO, Core Web Vitals and GEO readiness audit (100 pages Starter, 500 Growth and Agency, 5,000 Enterprise) |
| Checklist | Whatever you write down afterwards | One prioritised list scored impact x 2 minus effort, with how-to-fix guidance and copyable code |
| White-label | Not applicable | Agency plan (£399/month) and Enterprise: logo and colours, client workspaces, client-viewer logins |
| Price from | Free, plus your staff time | £49/month ex VAT, 14-day trial with card, two months free on annual billing |
A note on method: Seoptist collects answers only through official APIs (OpenAI, Gemini, Perplexity Sonar, Anthropic) and licensed data providers (SerpApi, DataForSEO). It never scrapes consumer interfaces with fake accounts, so the data is stable, permitted and free of your own chat history.
Who should choose which
Stick with manual checks if you want a one-off picture, are not yet sure AI assistants matter for your category, or just need to confirm one specific fact after a change. There is no reason to pay before you have done that.
Move to tracked, sampled measurement when you want to know whether you are improving, when you care about more than one engine, when wrong facts in answers would cost you enquiries, or when more than one person needs to see the numbers. The difference is not that the tool asks better questions than you can; it is that it asks the same questions, the same way, every week, and turns the answers into a trend and a to-do list.
If you want a middle step, the free AI Visibility Check at /free-check runs five prompts across two engines for your domain and sector. If that shows a gap worth closing, plans and what each includes are at /pricing.