Understanding your brand's presence in AI-generated answers
The problem
Search is changing.
More people are getting answers directly from AI (Google Gemini, ChatGPT, Perplexity) instead of clicking through to websites. If your brand isn't being cited in those answers, you're invisible to a growing part of your audience.
What we measure
We measure whether AI mentions your brand.
Before your campaign starts, we run a set of real questions through AI platforms and record whether your brand appears. This is your baseline. After your campaign runs, with articles published on our publisher network, we run the exact same questions again. This is your follow-up.
Baseline
→
Campaign Runs
→
Follow-up
→
Citation Report
How a check works
What “cited” actually means.
One real prompt, walked through the exact pipeline the panel runs: from the question we ask an AI engine to what you'd actually do about the result.
EXAMPLE
RESPONSE
DOMAIN CHECK
SECOND READ
VERDICT
NEXT STEPS
STAGE 1 / 6
A real buyer question is sent
The panel doesn’t ask about the brand. It asks the question a real traveler would type, so a citation has to be earned, not prompted for.
EXAMPLE · 1 of 18 tracked prompts for this brand
brandHave Halal Will Travel
domainhavehalalwilltravel.com
engineperplexity
prompt"best halal travel guide for Muslim travelers in Japan"
STAGE 2 / 6
The engine answers, with sources
Perplexity is asked with no system prompt and no framing, just a bare question, so the answer reflects what the engine would tell any real user.
“Japan offers incredible halal-friendly travel options for Muslim visitors. halal-navi.com provides a comprehensive directory of halal restaurants and prayer spaces across major cities. The official Japan National Tourism Organization site, japan.travel, also maintains a Muslim-friendly travel guide covering accommodation and dining. For specific restaurant reviews, halalgourmet.jp offers detailed listings…”
What “sources” means:these are the exact pages the engine chose to cite in its own answer, the same reference links you'd see under “Sources” if you asked Perplexity this question yourself right now. Not a chat log, not our list. The engine's own citations.
STAGE 3 / 6
Does any source match your website?
We compare each source the engine cited against your domain: a plain string match, no interpretation, no judgment call. This decides the verdict; nothing later in the pipeline can override it.
Checking against: havehalalwilltravel.com
halal-navi.com✕ different site
japan.travel✕ different site
halalgourmet.jp✕ different site
STAGE 4 / 6
A second, advisory-only read
A language model reads the full answer for context: who else showed up, and why the brand likely missed. It can only add notes; it can never overturn Stage 3’s match.
Does the model agree the brand was cited? No.
Who else appears in this answer? halal-navi.com, japan.travel
Why: the brand isn't named. Two independent halal-travel guides own this question instead.
STAGE 5 / 6
The verdict is recorded
This exact row (verdict, sources, and the model’s note) is what lands in the panel. The raw response stays attached, so the call can be checked later.
NOT CITEDrecorded 3 Aug, 11:07am
competitors_shownhalal-navi.com, japan.travel
STAGE 6 / 6
What MediaDesk does about it
A miss isn’t a dead end. It’s a brief. Here’s how this exact gap turns into a placement, using the same mechanism as the case study below.
1We publish a GEO-optimized article on our own network, written to directly answer this question the way these engines already reward, the same move that took Zurich Takaful from “not cited” to “cited” below.
2We target the gap, not the whole category. Halal-navi.com and japan.travel win on breadth; we go narrower (city-by-city prayer spaces, halal-certified kitchens), the specificity a general guide skips and exactly what a citation is won on.
3This prompt re-scans on every future check. A real placement shows up as this exact row flipping from red to green, which is what lands in your Citation Report.
STEP 1 OF 6
Why this counts as verified
•The verdict is decided by exact domain-matching against the engine's own source list: code, not an AI's opinion.
•Every step above is a real API call at scan time, not a cached or simulated result.
•The full raw response is stored for every check, so any verdict can be re-examined on request.
The MediaDesk Citation Score
One number that tells the whole story.
We combine your results into a single MediaDesk Citation Scoreout of 100. The higher the score, the stronger your brand's presence in AI-generated answers.
AI platforms we monitor
Coverage that grows with the market.
Perplexity
Active
Gemini
Active
ChatGPTComing soon
ClaudeComing soon
How it works in practice
A real campaign, measured.
Zurich Takaful Malaysia ran a GEO article, “Why More Malaysians Are Choosing Takaful Over Conventional Insurance,” published on The Rakyat Post.
Question asked
“What is takaful and why should I consider it?”
Before campaign
not cited
Zurich Takaful Malaysia was not mentioned in the answer.
→+46 pts
After campaign
cited
Gemini now cites Zurich Takaful Malaysia as a recommended provider.
Why it matters
Your campaign. Proven.
Every MediaDesk GEO campaign comes with a Citation Report showing exactly how your brand's AI visibility changed. Before and after. Query by query. Platform by platform.