Brand24, Mention and Meltwater all advertise AI citation monitoring features, but if you pay for all three you are probably paying twice for promises that haven't been kept yet. Only one of these tools has something usable today — the other two are on the roadmap. Knowing which one really works, tested on real cases, saves you months of useless subscriptions while you try to figure out what to do on your own.
Brand24, Mention and Meltwater are all adding “AI citation tracking” to their suites. One of the three already has a useful feature for measuring your visibility in AI answers. The other two, as of today, are roadmap promises dressed up as functionality.
Let me explain what I found by testing them side by side, why mention monitoring remains a central tool even when it isn’t yet mature on AI citations, and how to use it well while the market settles.
Why a boatyard in San Benedetto del Tronto should care about this
Think of a boatyard in San Benedetto del Tronto (AP) that builds inflatable boats and recreational craft under 10 metres. Niche market, long purchase decision, a customer who compares 4-5 Italian manufacturers before booking a sea trial.
If you open Perplexity today and ask “best Italian builders of inflatable boats under 10 metres”, you get a list of 5-7 names with citations. Those names aren’t there by chance: the AI chose them because it found converging mentions on boating forums, reviews, trade magazines and skippers’ blogs.
The point is obvious but it escapes anyone who doesn’t work on AI visibility: if you don’t know what the web is saying about your brand, you don’t know what the AI is learning about you. Mention monitoring is the radar that tells you which signals are reaching the models — even before you do or don’t appear in an answer.
AI citation tracking: why it’s a deductive mechanism, not yet a measurable factor
There is no Tier 1 paper today proving “Brand24 predicts your AI citations with accuracy X”. The claim here is deductive in nature, and I want to be honest about how it’s built.
In the world of language model research, the documented principle is that RAG (retrieval-augmented generation) systems like Perplexity and ChatGPT Search pull from the web in real time, while base models learned from the mentions accumulated in their training data. From this it follows that every new mention of your brand is a potential future signal: either it goes directly into the live grounding, or it contributes to the pattern the model will associate with your name when it uses its next version.
Translated into practice for the San Benedetto boatyard: a review on a skippers’ forum or an article in an online boating magazine can become, months later, one of the sources Perplexity will cite when a prospective customer in Trieste searches for “reliable 8 metre inboard inflatable”. Without monitoring you don’t know when the mention arrives, where from, and with what sentiment — so you can neither amplify it nor correct it.
This connects to what I’ve already explained in the articles on backlinks as a citation proxy and on implicit reference weight: the AI doesn’t distinguish between a “link” and a “textual mention without a link”, it counts both as signals of existence and authority.
Configuring only the exact brand name.
The test I ran on the three tools
I took the brand of a boatyard in the Marche region (it gave me consent, but stays anonymous) and configured Brand24, Mention and Meltwater in parallel for 30 days. Same brand, same keywords, same variants, same period. An indicative test, not a study: a single brand as a sample, but the pattern across the three tools was clear.
Tool A — Brand24: it captured 47 mentions in the month, 11 of them from boating forums, 8 from blogs/online magazines, 3 from social. Sentiment classified automatically (with some errors detecting irony), source filter working. The advertised “AI mentions” feature only shows the mentions that surfaced in ChatGPT and Perplexity after you’ve searched for them manually: it’s not proactive tracking, it’s an on-demand check. Useful but not automatic.
Tool B — Mention: 39 mentions captured, weak forum coverage (it missed 5 that Brand24 had found), good on news and social. “AI tracking” function announced on the roadmap but not available on the plan tested. A promise, not a tool.
Tool C — Meltwater: 52 mentions captured (the widest coverage, including transcribed podcasts), but a significantly higher cost and an interface designed for enterprise press offices. The “Generative AI Visibility” section shows aggregated data on the most widely used models, updated weekly. More mature than its competitors on the AI side, less accessible budget-wise for an SME.
Operational conclusion for an Italian SME: Brand24 remains the best value for money for classic mention monitoring, even though its “AI” feature is today more marketing than substance. For real AI citation tracking you still need dedicated professional tools or recurring manual checks on the AI engines.
Open a trial of Brand24 or Mention (both have 14 free days) and configure: exact name + 3-5 variants + the names of your best-selling models/products.
The mistakes I see most often when setting up a monitoring tool
Configuring only the exact brand name. The San Benedetto boatyard is called “X Marine” but gets cited as “X”, “X boatyard”, “X inflatables”, “X boats”. If you don’t add all the variants you lose 30-40% of the real mentions.
Ignoring mentions without a link. A textual citation on a forum is worth as much as one with a link for AI models: it talks about you, even if it doesn’t link to you. The tools capture them all, it’s up to you to look at them.
Filtering only for positive sentiment. Neutral mentions (descriptive, technical reviews on a forum) are often the most powerful for AI visibility: the models weigh existence and context, not emotion.
Forgetting vertical forums. For boating, forums like those of inflatable-boat enthusiasts or professional skippers count for more than 10 generic articles. Check that your tool indexes them.
Confusing volume with signal quality. A week with 50 low-value social mentions can look like a positive spike, but if in the same period you lose a citation in an authoritative boating magazine you’ve worsened your profile in the eyes of the AI models. What counts is where the mention comes from, not just how many there are.
How this work ties in with what you’ve already done on authority
Mention monitoring doesn’t live in isolation. It rests on all the upstream work on recognising your brand as an entity: if Google and the other engines don’t understand that “X Marine” is a boatyard (and not, say, a chain of car washes), the mentions you capture in the tool don’t turn into useful signals for AI visibility.
That’s why, before investing time in monitoring, it’s worth checking the prerequisites I’ve already covered in this series: your E-E-A-T for AI profile, the presence of a clean entry in Google’s Knowledge Graph, and the recognition of your bylined authors as entities (author entity recognition). Without these prerequisites, even 200 mentions a month won’t make you appear in AI answers: the model doesn’t know enough about you to trust you.
When the prerequisites are in place, though, monitoring becomes the lens that tells you which lever is working. You see a spike in mentions after publishing a technical white paper? That format works, you make more. You see that 70% of the quarter’s mentions come from three specific forums? You know where you need to be present as a brand, not just as a product.
What to do concretely this week
- Open a trial of Brand24 or Mention (both have 14 free days) and configure: exact name + 3-5 variants + the names of your best-selling models/products.
- Set a daily email alert for new mentions, a weekly one for the summary.
- Manually classify the first 30 mentions by: source type (media/blog/forum/social), context (descriptive/comparative/review), presence of a link.
- Compare with the 3-5 competitors the AI cites in your sector: search Perplexity for “best [your product] in Italy” and set up monitoring on their names too. Seeing what is written about them tells you where you should be getting noticed yourself.
- Once a month do a direct check on the AI engines (ChatGPT, Perplexity, Gemini): real queries from your customer, verify whether you appear, screenshot the cited sources. This is the real “AI citation tracking” today: partly automatic, partly manual.
Don’t expect the tool alone to tell you “you’re winning in AI answers”. It’s not mature enough for that yet. But it gives you the base signal — how many voices are talking about you, where, in what context — without which no AI visibility strategy has any foundation.
The thread: monitoring mentions is the radar, not the course
Measuring visibility in AI answers is layered work. Mention monitoring is the base layer: it tells you which signals exist in the world. In the following articles in the series I explain how to go from the radar to the course — how to turn the captured mentions into concrete actions that actually move your presence in ChatGPT, Perplexity and Gemini.
See also, in this series, the deep dives on how to measure share of AI voice, on how to track Perplexity citations manually, and on the reporting framework for AI visibility.
