What "free" honestly means here
The free AI visibility check in this guide is a manual method, not a software tool. You do the asking. You do the writing down. It costs nothing but about twenty minutes, and it tells you something real: whether the assistants your customers use name your business when they ask the questions your customers ask.
That is worth saying plainly, because search results for a free AI visibility tool mostly send you to one of two things. Either a sign-up form that gives you a teaser score and then asks for a card, or a page that promises a result and delivers a single prompt run once. A manual check is less glamorous than either, and more honest than both.
If you want the longer story of what software checkers do and where they fall short, our guide to AI visibility checkers covers it. This page does one narrower job. It walks you through the free method from the first question to the last row of notes, so you can finish it today.
Write ten real questions. Ask four assistants each one, logged out. Record who is named and which sources are shown. Read the pattern across all of it, not any single answer.
Who a free check suits, and who it does not
A free manual check suits you if you want a first honest look, you have a small number of things you sell, and you can spare an afternoon once a month. It is a good fit for a local business, a single store, a consultant, or anyone deciding whether AI answers are worth caring about yet.
It suits you less if you sell hundreds of different things, if you cover many cities, or if you need a weekly record across many questions. Twenty questions across four assistants is eighty answers. Doing that by hand every week turns into a part-time job, and people stop. A check that nobody runs is worth nothing, so be realistic about your own patience.
There is a middle path. Run the free version once to learn what a result looks like. Keep it going monthly on a small set. Only pay for software, or for a service, when the set grows beyond what you will actually keep up. Our page on AI visibility tools describes the categories you would be choosing between at that point.
One more thing about fit. A free check measures you against the questions you chose. It cannot tell you about questions you did not think of. Treat your first run as a map of a small area, not the whole country.
The twenty minutes, laid out
Here is how the time breaks down for a first run with ten questions and four assistants. Faster is possible once you have done it once.
- Five minutes: choose and write the ten questions.
- Two minutes: open a private window and the four assistants, and set up a notes page.
- Ten minutes: ask forty questions in total (ten each, four assistants) and jot down who is named. This is the longest part.
- Three minutes: count and read the pattern.
The figure of twenty minutes assumes you already know your category and your buyers, which you do. It does not include deciding what to do about the result. That is a separate conversation, and we give it its own page: the AI visibility checklist is the place to go once you know where you stand.
The diagram above shows the five steps in order. The important thing it marks is where checks tend to fail. They rarely fail at the reading. They fail at the start, when the questions are written in the owner's language, and in the middle, when the assistant already knows who you are.
Step one: write ten questions a buyer would type
This step decides the quality of everything after it. A check is only as honest as its questions.
Write them the way a stranger would say them to a friend. Not "best commercial roofing contractor Tampa" in keyword form, and not your brand name followed by a question mark. Something closer to "who should I call to fix a leaking flat roof on a small warehouse in Tampa." Ten questions is enough for a first run. If you can only think of six, use six.
A useful spread is:
- Three questions where the buyer knows the category and wants a shortlist ("best X for Y").
- Three where the buyer has a problem and does not know what to buy ("my X keeps doing Z, what should I do").
- Two comparisons ("X versus Y for a small team").
- Two with a place or a price limit in them.
Leave your own name out of every one. If the buyer already knows your name, they do not need an assistant to find you. The whole point is to see whether you appear when they do not.
The writing of a good set has enough nuance to deserve its own page. If you want to go deeper, read how to write the questions you test with. For a first free run, the spread above is enough.
Step two: ask four assistants, logged out
Use a private or incognito window for each assistant. Logged out is the standard, because a logged-in assistant can carry memory of earlier chats, and it may already know who you are and what you sell. You are trying to see what a stranger sees.
The four to start with are ChatGPT, Perplexity, Claude and Gemini. They are built differently, draw on different sources, and often disagree. Each one is a separate reading. If one name keeps appearing in all four, that is a stronger signal than the same name in only one.
Some assistants limit what logged-out users can do, or ask you to sign in after a few questions. If that happens, note it and carry on with what you can reach. Do not sign in with your business account and then pretend nothing changed. If you must sign in, say so in your notes, because the result is a different kind of reading.
Paste each question exactly as you wrote it. Do not rephrase halfway through to coax a better answer. The temptation to nudge ("and what about local options?") is strong, and it is the quickest way to measure your own prompting instead of the assistant. Ask once. Save follow-ups for later, after the first pass is recorded.
Google also offers an AI answer in its own results. That is a different surface with its own quirks, so it gets its own page: checking Google AI Mode.
Step three: write down what you see as you go
Do not rely on memory. The notes page can be a plain document or a sheet. For each question and each assistant, write four things.
- Were you named? Yes or no.
- How were you named? First, in a list of several, or mentioned in passing. Copy the exact words if you were named.
- Who else was named? Every business, in the order given.
- Which sources were shown? Perplexity and some other assistants list the pages they drew on. Write down the site names.
The third and fourth items are the ones people skip, and they are the ones that make a check useful. Knowing you were absent tells you very little. Knowing who was there in your place, and which pages the assistant leaned on, tells you where to look next.
If you want a ready-made layout for these columns, we describe one in an AI visibility tracking spreadsheet. For a first run, a blank page with a line per answer is fine.
Write quickly and do not editorialize. The goal at this stage is to capture, not to judge. You can feel bad about the results afterwards.
Step four: read the pattern, not the answer
Now count. For each assistant, how many of your ten questions named you at all? Across all four, how many of the forty answers named you? Then do the same for the two or three businesses that appeared most often.
Three simple readings come out of that.
You are absent almost everywhere. The assistants do not currently connect your business to these questions. That is common, and it is the most useful result to have, because it is clear where to start.
You appear in some assistants and not others. That points to differences in what each one draws on. A name that shows in Perplexity, which cites live pages, but not in ChatGPT, tells you something different from the reverse.
You appear, but others appear more. That is a competitive position, not a visibility problem. The businesses named in your place are your real rivals in this setting, and they may not be the ones you compete with in a shop window or on a Google results page.
What each of these readings means, and what to do about it, is the work of reading your AI visibility results and, after that, what to do after the check. Keep this page for the doing. Those pages are for the interpreting.
Why one run is a snapshot, and how to treat it
Assistants do not give identical answers each time. Ask the same question twice and the list of names can shift, the order can change, and one business can drop out while another comes in. That is how these systems work, not a fault in your method.
So treat your first run as a snapshot. It shows roughly where you stand, and which names recur. It does not prove you are invisible, and it does not prove you are safe. A name that appears in nearly every answer across several questions is probably reliably known. A name that appears once is probably luck.
If a particular answer matters to you, ask the same question two or three more times in fresh windows and see whether the result holds. Our explanation of why AI answers change goes into the reasons and suggests how many repeats make a fair reading.
The practical rule: do not decide anything big from a single run. Decide to run it again.
Where a free manual check falls short
It is worth being straight about the limits, so you do not over-trust the result.
Small sample. Ten questions and one answer each is a thin slice. It is enough to see a pattern, not enough to measure a small change.
Your location and device shape some answers. Local questions especially depend on where the assistant thinks you are. Run them from the place your customers are, or name the place in the question.
You are not a typical user. You know the topic well, so you may write questions that are too precise. Ask a friend or a customer to suggest three.
Nothing is stored for you. Software can keep a history. By hand, history exists only if you save your notes in the same place each time. That is why a fixed layout matters, and why we suggest one.
It does not tell you why. A check says what the assistants said. It does not by itself say what made them say it. For that you need to look at the sources behind the answers, which is part of the work covered in checking competitors in AI search.
None of that makes the free check pointless. It makes it a starting point, and honest about being one.
Turning a one-off into a monthly habit
The first run teaches you the most. The second run, a month later with the same ten questions, teaches you whether anything moved. By the third you have a trend, or at least an honest sense that nothing is changing.
To make that possible, three things need to stay fixed. The questions, the assistants, and the way you record. If you change the wording of a question, you have started a new series. Add new questions at the end if you want, but keep the original ten running unchanged beside them.
Put a reminder in your calendar for the same week each month. Do it at about the same time of day if you can. It sounds fussy, and it is, but a series is only comparable if you ran it the same way.
If you want a number to hold onto, the simplest honest one is "named in X of 40 answers". Resist inventing a clever score. Our page on the AI visibility score explains what a single number can and cannot carry.
A free check works if the questions are real, the windows are logged out, and the notes are kept in one place. Ten questions, four assistants, once a month.
Where RunOctopus fits, and where it does not
Be clear about what RunOctopus is and is not. It is not a free AI visibility checker, and it does not run this check for you on the free tier. The free site read at the link below does something else. You paste your website address, Ollie (our AI) reads the site, and it shows you the plan it would follow. It does not ask ChatGPT or any assistant what it says about you.
What RunOctopus does is the work that follows a check. It is a done-for-you service. Ollie researches your business, then writes and publishes pages directly on your own website, working with a team that plans, writes, reviews and publishes alongside it. It works with Shopify, WordPress, WooCommerce and Wix. There is nothing to install and nothing for you to approve, and the pages go live on their own.
Paying customers also get reports on where ChatGPT, Perplexity, Claude and Gemini name their business. So the manual check on this page is the free way to see where you stand today, and the reports are what you would have if you joined and wanted that watching done for you.
If you want to see what Ollie would build for your site, the free site read needs no account and no call.