Everyone is selling you an AI visibility dashboard. Google and Microsoft just gave you most of it for free.

There is a land grab happening in your inbox right now. Every week another platform emails you about "AI visibility" and quotes you $189 a month to tell you whether ChatGPT says your name.

Some of them are good products. One of them, Profound, raised $96 million on a $1 billion valuation in February 2026. That is real money chasing a real problem.

Here is the part they leave out of the sales email. In the last six months, Google, Microsoft and Google Analytics all shipped AI search reporting. For free. Built into tools you already have logins for.

So before you sign anything, spend an afternoon on the free stack. Then decide. That is what this post is.

What you are actually buying when you buy "AI visibility"

Strip the branding off and every one of these tools sells you one or more of four things:

• Presence. Does the model name you when someone asks a buying question?
• Citation. Does it link your page as a source?
• Traffic. Does anyone actually click through to you?
• Crawl. Are the AI bots even reading your site?

Four different questions. Most people mash them into one, then pay a vendor to half answer all four. Split them apart and you can see exactly which ones you already get for nothing.

Be honest about the order, too. Presence is worthless if the model has no idea who you are in the first place, and that is a different job entirely. We covered that in entity SEO. Tracking is the scoreboard. It is not the game.

The free stack: four reports that already exist

1. Google Search Console: Generative AI performance reports

On 3 June 2026 Google announced dedicated Generative AI performance reports inside Search Console. They split your impressions inside AI Overviews and AI Mode away from standard organic results, and do the same for Discover.

You get impressions, pages, countries, devices and change over time. You can filter it like any other Search Console report.

What you do not get, at least at launch: clicks, CTR, conversions, or the prompts people typed. Google is also rolling it out to a subset of site owners first, so if you cannot see it yet, that is why. It is not you.

Even with those gaps this is the most useful free datapoint on the list, because it is Google telling you directly which of your pages get pulled into its AI answers. No third party vendor has that data. They are modelling it. Google is measuring it.

2. Bing Webmaster Tools: the AI Performance report

Microsoft got there first. The AI Performance report landed on 11 February 2026 and was expanded on 16 June 2026 with intent labels, topic groupings, citation share and comparison tools.

It reports two things worth understanding:

• Grounding queries. The search phrases Copilot writes for itself when it decides it needs to go and fetch web content to answer someone.
• Citations. How many times a specific page of yours got used to build an answer.

Grounding queries are the interesting bit and almost nobody talks about them. They are the machine's version of keyword research. You get to see the actual questions the model asks the web on your behalf, which is a far better content brief than anything a keyword tool will hand you.

Scale check, and this is why you should not skip Bing. Otterly published its own numbers out of this report: across November 2025 to February 2026, 647 unique grounding queries pulled content from otterly.ai, producing 30,398 grounding events across 173 pages. That is a mountain of machine reading that shows up nowhere in a normal analytics report.

If you have never set up Bing Webmaster Tools because "Bing is small", you have been watching the wrong metric for two years. Copilot runs on it.

3. GA4: the AI Assistant channel

Google added the classification logic on 13 May 2026 and rolled it out widely on 7 June 2026. There is now a default channel group called AI Assistant. No setup, no regex, no custom channel group to build.

It pulls sessions from ChatGPT, Gemini, Copilot, Grok and DeepSeek out of the Referral bucket and gives them their own line.

Two honest caveats. It does not catch sessions that arrive with no referrer, and plenty of AI traffic arrives exactly like that, because people copy a link out of a chat window and paste it somewhere else. And the covered list is not every assistant on earth. Treat the number as a floor, not a total. Your real AI traffic is higher than what GA4 shows you. It is just unprovable, which is annoying but true.

4. Cloudflare AI Crawl Control

Formerly called AI Audit. If your site sits behind Cloudflare, this shows which AI crawlers are hitting you, how often, what they requested and whether they actually respect your robots.txt.

On the free plan detection works off user agent string matching, so it catches the bots that identify themselves honestly. Paid plans add behavioural fingerprinting for the ones that do not. You can also set per crawler allow and block rules, which is a decision worth thinking about properly rather than reflexively blocking everything and then wondering why you stopped getting cited.

One distinction that trips people up: crawling is not citation. A bot reading your page tells you that you are in the running. It does not tell you that you won. Early warning system, not scoreboard.

5. The prompt test nobody actually runs

Free, twenty minutes, and it is the only method here that answers the question you actually care about: does the machine say my name.

Write down the ten questions a real buyer types before they hire someone like you. Not keywords. Questions. "Who is the best damp proofing company in Newcastle." "Best CRM for a small plumbing business." Then run all ten through ChatGPT, Gemini, Perplexity and Copilot, and log who gets named.

Do it in a fresh session with memory and personalisation switched off, or you are just watching the model flatter you.

Keep it in a spreadsheet with a date column. Repeat monthly. That is a trend line, and a trend line is most of what a $189 dashboard is selling. We hand Dojo members the prompt sheet and the tracker already built so you are not starting from a blank grid, but you can absolutely make your own this afternoon.

The free stack at a glance

Build it in one afternoon

1. Verify Bing Webmaster Tools. Twenty minutes if you already have Search Console, because you can import the property straight across. Open the AI Performance report and export your grounding queries.
2. Open Search Console, Performance, Generative AI. If it is there, set the widest date range available and export. If it is not there yet, check back monthly.
3. Open GA4, Reports, Acquisition, Traffic acquisition. Look for the AI Assistant channel. Add a comparison against Organic Search so the number has context.
4. Switch on Cloudflare AI Crawl Control if you are on Cloudflare. Note which bots are heaviest and whether they are obeying robots.txt.
5. Build your ten prompt sheet and run it. Log the date, the model, the question, whether you were named, and who got named instead of you. That last column is the one that will annoy you into action.
6. Put a recurring hour in the diary, monthly. Done. This is the bit agencies quietly bill $2k a month for, and once the plumbing is in it is one hour, once a month.

Now you have four data sources covering crawl, citation, presence and traffic, and you have spent nothing.

Where the free stack genuinely stops

I am not going to pretend free does everything. It does not. The honest list:

• Competitors. Search Console and Bing only report on your property. You cannot see who is beating you or how often. Paid tools can.
• Prompt level history at scale. Ten prompts by hand is fine. Four hundred prompts across six engines and forty countries every week is not.
• ChatGPT and Perplexity citation data. Nobody hands you a first party report for these. Google reports on Google. Microsoft reports on Microsoft. OpenAI reports on nothing.
• Sentiment. Being mentioned and being recommended are different outcomes. The free stack cannot tell them apart.
• Multi brand reporting. Run twelve clients and manual tracking collapses fast.

That gap is exactly what the vendors sell into. Fair enough. The only question is whether it is your gap.

What the paid tools cost, and who they suit

Published pricing as of writing. Vendors move this often, so check before you buy anything.

Rough guide to what you should expect to spend, by size:

• Local business, one location: $0. The free stack plus a monthly prompt sheet is genuinely enough. Put the budget into local SEO instead, where it moves something.
• Growing service business, one brand: $0 to $50 a month. A cheap tier buys you automation, not new information.
• Agency, five clients or more: $100 to $250 a month. Here the maths works, because the reporting time you save is real money.
• Enterprise: custom. If you have a legal team that cares what an LLM says about your brand, you already know why you are paying.

Who should not buy a tool yet

Blunt version. Do not buy an AI visibility tool if:

• You have never once run your buying questions through ChatGPT by hand. Do the free version first. If you are nowhere, a dashboard just tells you that in colour, monthly, for $189.
• You have no plan for what you would do with the data. A tool that reports you are invisible does not make you visible. That is a content and mentions problem, not a reporting problem.
• You have one brand, one market, one language. You are paying for scale you do not have.
• Your site has fewer than thirty pages worth citing. Fix the pages first.

Buy one if you are an agency, if you are in a knife fight of a category and need to watch rivals, or if a board wants a number every month and "I checked by hand" will not fly.

The mistakes I see every week

Treating a crawl as a win. GPTBot hitting your site 4,000 times means the machine read you. It does not mean it repeats you. Different metric, different job.

Chasing mentions instead of recommendations. Getting listed under "other options include" is not the same as being the answer. Log the position, not just the presence.

Testing while logged in. Memory and personalisation will inflate your results and make you feel great about nothing. Fresh session, every time.

Buying a tool to fix a content problem. The most common one by miles. If AI has never heard of you, no dashboard changes that. Third party mentions do. Start with getting AI to recommend your business, then measure.

Ignoring Bing because it is small. Copilot is not small.

Expecting clicks. A large share of AI answers end the journey right there. Impressions inside an AI answer are brand exposure, not a traffic channel. Judge it on the right axis or you will talk yourself out of something that is working. Same argument applies to Google AI Mode.

Alternatives if you hate dashboards entirely

Two that cost nothing and work:

Server log review. Not on Cloudflare? Your raw access logs already contain every AI crawler hit you have ever had. Filter for the known user agents once a month. Uglier than a dashboard, same information, arguably more accurate.

A saved prompt set inside each model. Keep one chat per engine with your ten questions pasted in. Re run it, screenshot, date it, drop it in a folder. Crude, and completely effective. Plenty of our members do exactly this and never buy a tool.

Risks worth naming

The reporting is new and it moves. Google is still rolling its report out and has said the data will develop. Bing has already expanded its report once this year. Anything you build in August 2026 should get reviewed in six months, and any vendor promising a stable long term benchmark is overselling.

Second risk, and the bigger one: measurement theatre. It is very easy to spend six months building a beautiful AI visibility dashboard and zero months building the third party mentions, the citations and the answer first pages that actually get you named. Measure lightly. Build heavily.

Where this goes by the end of 2026

My read, stated as opinion and not fact. First party reporting keeps improving, because it is in Google and Microsoft's interest to keep publishers pointed at their own ecosystems. Clicks and query data inside the Search Console AI report are the obvious next shoe to drop.

The moment that happens, the bottom half of the paid tool market gets squeezed hard, because "we show you your Google AI impressions" stops being a product. What survives is the stuff the platforms will never give you: competitor tracking, sentiment, and coverage of the engines that report nothing at all.

So if you are buying now, buy for those three things. Not for a number Google is about to hand you for free.

The move

Spend one afternoon this week. Verify Bing, check Search Console, look at your GA4 AI Assistant line, run your ten prompts. Write down what you find.

Then, and only then, decide whether a tool earns its money. Most of you will land on "not yet", and the real work is on the page and off the site, not in the dashboard.

That work is what we run inside the Dojo: the prompt tracker, the answer first page templates, the mentions playbook, the tool stack, and the order to do it all in. If you would rather we just did it for you, that is Done For You. If you want it built into a proper search programme, start with our SEO service.

Join The Dojo. $67/mo. Get in here.