presence
Mention frequency
How often, across the frozen prompt set, each engine names you at all — the base metric everything else is read against. A falling frequency is the earliest signal of drift, usually visible before anyone in sales feels it.
managed ai monitoring
AI visibility monitoring, run as a service — not another dashboard.
The market is full of tools that will tell you your brand slipped out of ChatGPT this week. Almost none of them do anything about it. We run the watching and the acting: a fixed prompt set tracked weekly across ChatGPT, Claude, Gemini, Perplexity and Google's AI answers, drift triaged the week it appears, and the corrections we can make at source made for you. Tool-agnostic, bilingual, with a monthly price you can read below — because a citation you won is a position you have to defend.
why it matters now
Citations decay, and the loss is invisible until it costs you
AI answers have a property classic search never had: they are unstable from week to week. Tracking across 2026 finds only around a third of brands visible in one AI answer remain visible in the next, as models refresh, re-retrieve, and reweigh their sources. A brand that owned its category in AI answers in spring can be absent by summer, not because anything on its own site broke, but because a competitor earned a better mention or an engine started leaning on a source that does not name you.
The loss is quiet, which is what makes it dangerous. There is no ranking drop in a dashboard you already watch, no traffic cliff to investigate, because most of these answers never produce a click in the first place — by recent estimates the large majority of AI-mode sessions end without anyone visiting a site at all. The impression happens inside the answer, and if your name is not in it, you simply were not considered. Teams discover this a quarter late, from a pipeline that thinned for reasons the analytics cannot explain.
And the surface is growing, not shrinking. Estimates of how often Google shows an AI Overview vary widely by method and query type, and should be read with caution, but the direction is not in dispute: one 2026 tracking reading put Overviews on roughly half of searches, up sharply from a few months earlier, and ChatGPT alone fields billions of prompts a day. Whatever the exact percentage, more of your buyers' research now resolves inside an answer than last year, which means more of your visibility is decided in a place a traditional analytics stack cannot see. Monitoring is how you get eyes on it.
Monitoring is the instrument that turns that invisible loss into a number you can act on while it is still small. It is also, bluntly, the step most programs skip — they treat AI visibility as a project that ends rather than a position that has to be held. It is step six of the six-step method for a reason: everything earned in the first five steps decays without it.
the gap every tool leaves
A dashboard measures the problem. It does not solve it.
There are good AI visibility tools — Profound, Peec, Semrush's AI toolkit and a dozen others — and they do real work surfacing mentions, share and sources. But a tool hands you data and stops. It cannot notice that the reason you lost a citation is a two-year-old pricing figure on a comparison site, draft the correction, and get it changed. It cannot tell that a new Reddit thread is now feeding three engines and decide whether you should be present in it. That judgment, and the doing, is the part that moves your numbers, and it is exactly the part no dashboard includes.
The predictable result is the graveyard of unread dashboards: a subscription bought in a burst of urgency, checked twice, and then ignored while the very drift it was meant to catch happens unwatched. The data was never the missing piece. The missing piece was someone whose job is to look every week and act on what they see. That is what this service is, and it is why it is priced as a service rather than a seat.
It is also why we stay tool-agnostic. We are not selling you our software, so we will run your monitoring on whatever fits — an enterprise platform where you need depth and procurement, a leaner setup where you do not, or our own tracking where that serves best. You are never locked into a dashboard whose main feature, from your side, is a monthly invoice.
There is a practical edge to that independence beyond price. When the monitoring lives with the people acting on it rather than inside a vendor's platform, your history and your prompt set are yours: if you ever take the work in-house or move it elsewhere, the baseline and the trend line come with you instead of staying trapped behind a cancelled subscription. The data should serve the decision, and the decision should belong to you — not to whichever tool happened to be open when the number changed.
what we track
Four signals, every engine, every week
share
Share of answer
Of the brands an engine names for a category prompt, what fraction is you versus each competitor. Share of answer is the AI-era equivalent of share of voice, and it moves when a rival earns a mention you did not.
accuracy
Accuracy & sentiment
Not just whether you appear, but whether the description is right and the framing is favorable. A confident, wrong answer about your pricing or your category does more damage than silence, and it is the kind of problem a number-only dashboard hides.
sources
Cited sources
Which URLs and domains each engine leans on to build its answer for your category. This is the early-warning system for everything else: when a new source starts feeding the engines, you want to know before your competitor is the only brand named in it.
prompt set frozen for comparability · ChatGPT · Claude · Gemini · Perplexity · Google AI Overviews
the weekly loop
Detect, triage, act, report
01
Detect
The weekly run surfaces every change against last week and against your baseline: a lost citation, a new competitor in the mix, a description that drifted inaccurate, a fresh source the engines started trusting.
02
Triage
We sort what we see by impact and by whether it is fixable here or needs escalation. A wrong pricing line repeated by three engines jumps the queue; a one-off blip on a low-value prompt waits and is watched.
03
Act
Source-level corrections, content refreshes on slipping pages, and handoffs to entity or earned-media work where the fix is structural. The remediation we can do directly is included; the larger plays are scoped and quoted, never silently billed.
04
Report
A monthly read of what changed, what we did, and what moved — per engine, against the baseline. Short enough to actually read, honest enough that a flat month is shown as a flat month.
The loop is what separates monitoring-as-a-service from monitoring-as-a-feature. Detection alone is a notification; the value is in the three steps after it. Most weeks the work is small and quiet — a correction here, a refresh there — which is exactly the point. Held steadily, small weekly interventions keep a citation position that would otherwise erode. Skipped for a quarter, the same position has to be rebuilt from scratch, at the cost of the earned-media and entity work that is far more expensive than the maintenance would have been.
what a month actually looks like
Most weeks are quiet. The quiet is the product.
It helps to be concrete about what you are paying for, because "monitoring" sounds like a lot of dramatic dashboards and is mostly the opposite. A typical week is fifteen minutes of triage on a run that found nothing urgent: your numbers held, the descriptions are still accurate, no new source shifted the picture. That non-event is the service working. The drift that did not turn into a loss is invisible by definition, which is exactly why teams undervalue maintenance until they stop doing it.
Then, every few weeks, something moves and the loop earns its keep. A comparison site updates and suddenly describes you in the wrong category; we catch it on the next run and get the correction filed before three engines have repeated it. A competitor lands a mention in a source the engines trust and starts taking share of answer on a money prompt; we flag it, assess whether you can be present in the same place, and route it to earned-media work if so. An engine refreshes and your best-performing page slips; we refresh the content before the slip compounds. None of these is visible in a tool until after it has cost you, and none of them fixes itself.
At the end of the month you get a report built to be read in five minutes: what changed, what we did, what moved, per engine, against the baseline. A flat month is reported as flat — we would rather show you an honest plateau than manufacture activity, because the point of monitoring is an early warning you can trust, and a warning system that cries wolf is worse than none.
where it sits
Monitoring is the feedback loop for everything else
On its own, monitoring defends a position. Wired into the rest of a GEO program, it becomes the instrument that tells every other service where to aim. The same baseline our AI visibility audit establishes is what monitoring tracks forward. When it detects an entity being misread, the fix routes to entity and schema work; when it detects a source the engines trust where you are absent, that becomes a target for citation engineering. Without monitoring, those services run on last quarter's picture; with it, they run on this week's.
This is why monitoring is the natural standing layer of a full program rather than a one-off. The audit is a snapshot, the sprints are projects, but visibility is a moving target — and the thing that keeps the projects aimed at where the target actually is, week after week, is the monitoring loop. The AC Group has watched and adjusted to shifting attention for 27 years; the surfaces changed, the discipline of watching closely did not.
The other quiet benefit is that the record compounds. A month of monitoring is a status check; a year of it is an asset — a dated history of how each engine's treatment of your brand and your category moved, which sources rose and fell, and which interventions actually shifted the numbers. That history makes every future decision sharper: you stop guessing whether a tactic works because you can see what happened the last three times something like it was tried. Most brands never build this record because they treat AI visibility as a series of disconnected projects. Run as a standing loop, it accumulates into the one thing a fast-moving channel makes hardest to get — memory.
fit, honestly
Who this is for, and who should wait
Managed monitoring fits brands that already appear in AI answers, at least sometimes, and cannot afford to lose that ground — companies far enough into AI visibility that defending it is now the priority. It is also the right standing service after an audit and a sprint or two, the layer that protects the investment those projects made.
Who should wait: if engines rarely or never name you yet, monitoring will mostly confirm an absence you already suspect, and your money is better spent earning the first citations than watching for ones that are not there. Start with the audit to size the gap, then the work to close it; monitoring becomes worth its monthly cost once there is something to protect. We will tell you which side of that line you are on before you sign anything — a monitoring retainer sold to a brand with nothing yet to monitor is the kind of easy revenue we would rather not take, because it does not survive the first quarterly review.
There is a second group this fits especially well: teams that have a tool already and are not using it. Buying the dashboard was the easy part; the weekly discipline of reading it, interpreting drift and acting is what fell off the calendar three weeks in. If you have a Profound or a Peec seat gathering dust, the managed service is often the cheaper fix — we run the cadence on the platform you already pay for, and you get the acting layer the subscription was always missing. The software was never the bottleneck; the standing attention was.
pricing · no "contact us"
Monthly, fixed, published
managed monitoring
€1,500/month
One language. Weekly multi-engine tracking on a frozen prompt set, drift triage, the corrections we can make at source, and a monthly action report.
Start monitoringmanaged monitoring · plus
€2,600/month
Everything in Managed, plus competitor share-of-answer tracking, bilingual EN/ES coverage, and a standing remediation budget so most fixes happen without a separate quote.
Start with PlusMonth to month, cancel anytime — monitoring should keep earning its place, not lock you in. Tooling costs, where a third-party platform is the right fit, are passed through at cost and shown on the invoice; we do not mark up software. Remediation beyond the included scope is quoted before it happens, never after. And if a few months in the numbers are stable enough that a lighter touch makes sense, we will say so and step the engagement down rather than quietly keep billing the full rate — the same logic that makes us decline brands with nothing to monitor applies to clients who have grown steady enough not to need the heavier tier.
questions
Frequently asked questions
How is this different from an AI visibility tool like Profound or Peec?
Those are dashboards; this is a managed service. A tool shows you that your citation share dropped. We notice the drop, work out why, and do something about it — correct the source a model is misreading, refresh the page that lost its citation, or flag the competitor mention that needs answering. We are also tool-agnostic: if a platform like Profound or Peec fits your case we will run on it, and if a lightweight setup serves you better we will use that. You are paying for the watching and the acting, not for another login your team forgets to check.
Why do I need ongoing monitoring at all?
Because AI citations are volatile. Tracking across 2026 finds only around a third of brands visible in one AI answer remain visible in the next as models refresh and re-retrieve. A brand can dominate AI answers in March and quietly vanish by June because a competitor earned a better mention or an engine shifted toward different sources. Without monitoring you discover that loss a quarter late, from a sales pipeline that dried up; with it, you catch the drift the week it starts.
What exactly do you monitor?
A fixed set of buyer-relevant prompts, run weekly across ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews. For each we track whether you are named, your share of answer against competitors, the accuracy and sentiment of how you are described, and which sources the engines cite to build the answer. The prompt set is frozen so the trend line stays comparable over time, which is the part teams running it themselves most often get wrong.
Do you only report, or do you also fix things?
We act. Monitoring without a response is just anxiety on a schedule. When the weekly run surfaces a problem we triage it: source-level corrections where an engine is repeating a wrong description, content refreshes where a cited page is losing ground, and escalation to earned-media or entity work where the fix is bigger than a tweak. The monthly report shows what changed, what we did, and what moved as a result.
How quickly will I see drops or improvements?
Weekly monitoring means most drift is caught within seven days of it appearing. How fast a fix shows up depends on the engine: Perplexity and other retrieval-heavy systems can reflect a correction within days, while engines leaning on periodically refreshed knowledge take longer. We report on the cadence the engines actually move at, not a vanity daily number that mostly measures noise.
Can you monitor Spanish-language AI answers too?
Yes. The Plus tier runs the same monitoring across Spanish-language prompts and sources, with one unified view of your entity in both languages. For brands serving the US plus Latin America or Spain this matters more than it looks, because the two language ecosystems drift independently and a problem in one is invisible if you only watch the other.
See your starting position, free
The free AI visibility snapshot is the baseline monitoring tracks forward: how five engines name you today, and whether you have a position worth defending yet. It is the honest first step, whether or not monitoring turns out to be your next move.