AI visibility monitoring and tracking
Measuring the same questions on a schedule, so the answer is a trend rather than a screenshot. Including the thing most trackers are vague about, which is how often they actually ask.
AI visibility monitoring asks a fixed set of questions of each assistant on a schedule and records whether you were named. The point is not the number, it is the direction: generative answers vary between runs, so a single reading is one sample and only a repeated one tells you whether anything you did worked.
- The unit
- A rate across a fixed prompt set, per engine. Never a position, never one blended score.
- What it is for
- Direction. A single reading cannot tell you whether you are improving.
- Our cadence
- A scheduled read, monthly to daily depending on plan. Not a live ticker, and the plans say the run counts.
- What it cannot do
- Attribute revenue. Assistants pass almost no referrer data.
Why one reading is worth almost nothing
Ask an assistant the same question twice and you can get two different answers, with different sources, on a site that changed nothing in between. This is not a fault in the assistant or in the tool measuring it. It is how generated answers work, and every honest thing about monitoring follows from it.
So the screenshot doing the rounds in your industry group chat, showing a competitor named and you absent, is one sample. It might be the general case. It might be the one run in ten where that happened. From a single ask there is no way to tell, and the confident conclusions people draw from screenshots are the single most common error in this category.
What survives the variance is a rate. The same questions, asked of each engine separately, on a schedule, recorded the same way each time. Twenty questions asked monthly for six months is a finding. One question asked once is an anecdote with a timestamp.
Cadence, and the question to ask any vendor
This is the part of the category that is quietly vague, so here is ours plainly. We run a scheduled read, not a continuous one. The number of runs a month is a plan limit and it is printed on the pricing page rather than described as "real time" anywhere: the free tier asks 3 questions once a month, and the paid tiers move up through 4, 12 and 30 runs a month with more questions each.
That is a deliberate product decision rather than a limitation we are apologising for. Each run puts real questions to real assistants, which costs money per ask, and asking hourly would multiply the cost without improving the finding, because the thing being measured does not move hourly. What moves is your site, your competitors' pages, and the third party sources the engines draw on, and those move on a scale of weeks.
The question worth asking any vendor in this category is simply: how often do you actually ask, and how many questions. A product describing itself as real time monitoring is either asking constantly, which is expensive and shows up in the price, or it is running a schedule and choosing a more exciting word for it. Both are fine. Only one of them is being straight with you.
Per engine, never blended
An average across four assistants hides the only finding worth acting on, which is that one of them cannot see you. A brand strong in three engines and absent from a fourth has a specific, diagnosable problem, and a single score erases it.
Named and linked are different outcomes
Being written into the answer and being listed as a source are separate results with separate fixes. A tracker that reports only mentions is collapsing the two most useful states into one.
Record who was named instead
The competitor list is the most actionable line in any monitoring report, and it is the one people forward internally. It also tells you whether your problem is your own pages or the third party sources the engine preferred.
Keep the deterministic half beside it
Crawler access, render output and schema validity do not vary between runs. Tracking them alongside the sampled numbers is what lets you tell a real change from noise, because a drop that coincides with a robots.txt edit is not a mystery.
What to compare, if you are shopping
Entry prices in this category run from free to a few hundred a month, and the headline figure is the least informative part of the comparison. Four things decide what you are actually buying, and three of them are usually below the fold.
Engines covered at your tier, not on the pricing page overall. At least one product in this category sells a cheap plan that tracks a single assistant, with the comparable plan several times the price. Questions and runs, which is the resolution of the whole thing. Seats, because per seat pricing turns a cheap plan expensive the moment a second person needs to look. And history, since the entire value is the trend and a tool that keeps ninety days cannot show you a year.
We publish a comparison of the vendors in this category with prices read from their own pricing pages and their own sites scored by our engine, and a separate one that splits the work into the four jobs it consists of, because monitoring is only one of them.
What it costs
Every limit below is enforced by the server rather than advertised loosely, and the free tier is a real tier rather than a trial.
- 1 site, 20 pages a crawl
- 36 checks on every page
- 3 AI questions, asked once a month
- 1 keyword tracked weekly
- A free account and no card for your first audit
- 5 sites, 50 pages a crawl
- 25 keywords tracked weekly
- 10 AI questions, 4 runs a month
- 3 crawls a day, 45 a month
- PDF and CSV export, sent to any address
- Everything in Individual
- 20 sites, 100 pages a crawl
- 100 keywords tracked daily
- 25 AI questions, 12 runs a month
- 3 competitors watched per site
- Scheduled reports by email
- Everything in Startup
- 50 sites, 500 pages a crawl
- 500 keywords tracked daily
- 50 AI questions, 30 runs a month
- 5 competitors watched per site
- White label reports under your own brand
Annual billing is cheaper on every paid tier and the figures are on the pricing page, which is the authoritative source if anything here ever disagrees with it. Service plans, where our team ships the fixes rather than reporting them, start separately at $99 a month.
Questions people ask about AI visibility monitoring
What is AI visibility monitoring?
Asking a fixed set of questions of each AI assistant on a schedule and recording whether your brand is named, whether it is linked, and who is named instead. The output is a rate over time rather than a score, because generative answers vary between runs and only a repeated measurement can tell you whether something you changed made a difference.
How often should AI visibility be tracked?
Often enough to see a trend and no more. What is being measured moves on a scale of weeks, driven by your pages, your competitors and the third party sources engines draw on, so a monthly or weekly read is usually the right resolution. Hourly asking multiplies cost without improving the finding.
Is your tracking real time?
No, and we would rather say so than pick a more exciting word. We run a scheduled read, and the number of runs and questions is a published plan limit rather than an adjective. Each run puts real questions to real assistants, which has a cost per ask, and the thing being measured does not change fast enough to justify constant polling.
What is the difference between an AI visibility audit and monitoring?
Time. An audit is a photograph: what is true about your site and your presence right now, most of it verifiable by anyone. Monitoring is a film: the same questions asked repeatedly so you can see direction. Buy the audit first, because it is cheap and it finds the mechanical problems, and monitor once you have something worth watching.
Can I track AI visibility for free?
You can check it for free, which is not the same thing. A free check tells you where you stand today across a couple of assistants. Monitoring means the same questions asked the same way repeatedly with the history kept, and that is what paid tools in this category are actually selling. Start with the free check and only pay once it has shown you something.
Can monitoring tell me how much revenue AI search brought me?
Not honestly. Assistants pass very little referrer data, so sessions that began in an AI answer usually arrive looking like direct traffic. Any precise revenue figure attributed to AI visibility is a model, and a trustworthy version of that claim states its assumptions rather than presenting the output as a measurement.
Which engines should be monitored?
The ones your buyers use, which you find out by checking rather than assuming. ChatGPT is the most used and searches through Bing, which surprises sites that are strong on Google and thin on Bing. Claude fetches live and cites explicitly. Gemini tracks Google closely. Perplexity leans hardest on third party sources, which makes it the earliest warning that your corroboration work is or is not landing.
Get the photograph before you buy the film
The free audit covers the deterministic half on your live pages, which is where most problems turn out to be and which no amount of monitoring would have fixed. Free, no card, and the report is yours whether or not we ever speak.