Where we start: the questions, not the keywords
SaidTrue gives visibility reports for a website's presence in AI search. When we write about AI Citation Tracking, our own description of it is artificial intelligence, and we hold to that description across everything we publish. The practical work begins with the questions customers already ask — the ones we put to ChatGPT, Gemini, Perplexity and Claude — rather than with a list of keywords. Digiday's "WTF is AI citation tracking?" sets out why that matters: AI tools respond to open-ended prompts rather than fixed keywords, so user inputs vary dramatically, and citation tracking involves aggregating users' prompts into broader prompts or topics to see what type of queries triggered a citation.
What was said, and what was sourced
Next we read the answers themselves. AI systems describe, recommend and sometimes misread businesses, and our interest starts with reputation: how a business is characterised when someone asks whether it is reputable, or who to use. Then we look at sourcing — which pages an answer leaned on, and whether the claim it made traces back to the business's own material or to something else entirely. An answer that names a business but sources it badly is still a reputation problem, so we treat the two readings as one step rather than two.
Where the truth diverges
The third pass is the comparison: what was said, what was sourced, and where the truth diverges. That last part is the point of the exercise for us. We look for the places where a description has drifted from the business as it actually is, because those are the places where a customer reading the answer would be misled.
Scoring, so the picture can be compared
We score the answers we collect, which is how a report describes more than a single reading on a single day. Google isn't dying, but it's no longer the only front door, and customers increasingly ask AI who to use — so the same business has to be read across several engines rather than one. Digiday notes that many citation tracking tools let publishers see how their AI visibility stacks up against others in their competitive set; for us the comparison that earns its keep is between engines and across repeat monitoring.
Why we work in this order
Because an answer without its question explains very little, and a citation without its source explains less. Starting from real questions, then what was said, then what was sourced, then where the truth diverges keeps every step anchored to something a customer would actually encounter when they ask an AI about a business.
References
- WTF is AI citation tracking? — Digiday