See your AI visibility as one broken rung

AI visibility monitoring tracks how your brand appears in AI-generated answers across engines like ChatGPT, Perplexity, Gemini and Claude, and in Google AI Overviews. Over 70% of discovery queries now return a zero-click AI answer: category research, comparisons, recommendations. Where your brand appears in those answers, or fails to, directly affects traffic. TrustData probes ChatGPT and Perplexity daily on every plan, reads AI Overviews per keyword, and scores readiness on six independent rungs rather than one blended index.

Because an average hides the one thing you can act on. A brand can score well on crawling and still be invisible in answers, or be mentioned constantly and never referred. Averaging six signals into one number tells you that something is wrong without telling you what. TrustData scores each rung independently against a healthy cut-off of 60, then marks exactly one focus rung: the lowest rung below the cut-off whose prerequisites are already clear. Rungs above an unhealthy rung are shown as locked, because working on them before the lower rung is fixed does not produce a result. A rung with too little data reads as insufficient data and is excluded from focus selection rather than scored as zero.

Get Found (can AI crawlers reach and retrieve you), Be Understood (do machines resolve you to the right entity), Be Retrieved (does your content survive extraction), Be Trusted (is the framing around your brand positive), Be Chosen (do you get named on competitive questions), and Enable Transactions (does AI referred traffic behave). They are scored in that order because each depends on the ones below it. There is no point optimising sentiment on pages AI crawlers cannot reach.

AI Overview state is read per tracked keyword and reported as a rolling four-week presence rate with the observation count behind it, not as a yes or no. When an Overview appears we also record whether you were referenced, your position among the sources, and the snippet shown, then compute a citation rate over the same window. Presence tells you the surface exists for that query; citation rate tells you your share of it. All of it sits on the same keyword row as impressions, clicks, position and conversions, so you can see the trade between a lost blue link and a won citation. AI Overviews are generated, personalised and A/B bucketed, which is exactly why a single reading rendered as a boolean overstates what anyone actually knows.

TrustData probes 6 engines: ChatGPT, Perplexity, Gemini, Claude, Mistral and DeepSeek, and reads two Google SERP surfaces, AI Overviews and AI Mode. ChatGPT and Perplexity ship as baseline probe engines on every plan, alongside AI Overview tracking. Gemini (25 to 199 EUR), Claude (99 to 799 EUR), Mistral and DeepSeek (15 to 99 EUR each) are per-engine, per-tier monthly add-ons. Google AI Mode is the exception: it is read once per tracked keyword rather than probed, so it is priced per keyword at 29 to 299 EUR. There is no per-probe overage: engine cost is amortised into the add-on price and the scheduler caps at your tier's probe budget. Grok was dropped in July 2026 rather than left on the list, because it never returned a usable row.

The Brand Health Score is a 0 to 100 composite measuring how consistently your brand appears in AI answers. It is a weighted average of your with-search mention rate (60%) and no-search mention rate (40%) across all engines and question types, updated daily. It returns no value at all below 5 probes rather than reporting a number built on noise. It is a supporting metric. The readiness ladder, not this score, is how you decide what to do next.

When you follow a recommendation, TrustData records the change with a field-level diff and starts tracking the prompts it should have affected. Those are the treatment group. Your other prompts keep running unchanged on the same schedule, and their drift is the natural control. The verdict is the adjusted delta, treatment change minus control change, which filters out noise from model updates, competitor activity and seasonality. A winner verdict requires a 10 point adjusted delta after at least 7 days, a deliberately high bar given typical probe sample sizes.

We do not claim that, and we will not report it as if we could. TrustData runs two regimes of evidence and never bridges them. Paid media, audiences and smart links are measured causally through geo, time and platform holdout tests, reported with explicit confidence intervals. AI visibility, content and SEO are measured as leading indicators: citation rate, share of voice, ranking lift, each with an explicit noise level. The sixth rung reports AI Search sessions and their conversion rate as proxy signals, labelled emerging. Any tool telling you it has attributed revenue to a ChatGPT citation is overstating what the data can carry.

They are two different Google surfaces. An AI Overview is the generated summary that appears above the blue links on a normal results page, and TrustData reads it per prompt as a rolling four-week presence rate. AI Mode is a separate conversational surface with its own answer, its own references and its own cited domains, and TrustData reads it once per tracked keyword. Because they are read differently they are billed differently: AI Overview tracking is included in every tier, while AI Mode is priced per tracked keyword from 29 EUR. They are reported side by side and never blended into one Google score, because a brand can hold the AI Overview citation on a keyword and be absent from AI Mode on the same keyword.