Competitor discovery — rivals surface themselves
The competitor that matters is not the one on your pitch deck — it's the one AI keeps recommending to your buyers. Every scan reads its own answers and extracts the brands the engines actually named, so new rivals appear on your Competitors page the week they enter the conversation, with a count of how many answers they showed up in.
How it works
Extracted from the answers you paid for
The scan's stored answers are read by a judge model that pulls out every product recommended in your category — no extra scan spend, one cheap extraction call.
Folded so brands count once
“Copilot Pro”, “Copilot Business” and “GitHub Copilot” are one product. Alias resolution folds tiers, renames and spelling variants before anything is counted.
Promoted when you say so
Discovered brands appear with their answer counts — “named in 13 answers” — and one click adds them to your tracked rivals, where the next scan measures them properly.
In depth
A threshold against noise
A brand mentioned once in ninety answers is probably an aside. Discovery only surfaces brands named in two or more answers, so the list stays short and worth reading.
Tracked rivals get the full treatment
Once tracked, a rival appears in the visibility leaderboard with its named rate, margin, per-engine split and week-over-week delta — and in the framing panel with the verbatim phrases AI uses for it. Starter tracks 6 rivals, Growth and Scale 10.
Drops are information too
A rival fading out of answers is as much a signal as one arriving. Because discovery re-runs on every scan, the counts move week to week — and the leaderboard's deltas show who's climbing before it costs you.
What we won't do: auto-track whatever the extractor finds. Extraction is a model reading model output — good, not perfect — so discovered brands are suggestions with counts, and tracking stays your decision. Nothing enters your leaderboard without your say.
Common questions
How does CiteGraph find my competitors?+
By reading the scan's own answers: a judge model extracts every product the engines recommended in your category, folds duplicates and renames through alias resolution, and surfaces brands named in 2+ answers on your Competitors page with their counts. You choose which to track.
What happens when I track a discovered rival?+
The next scan measures them fully: named rate with margin, per-engine split, week-over-week delta in the visibility leaderboard, and the verbatim phrases AI uses to describe them in the framing panel.
How many competitors can I track?+
Starter tracks 6, Growth and Scale 10. Discovery itself is unlimited — it shows every brand that clears the threshold, tracked or not.
Does discovery cost extra?+
No. It reads answers the scan already collected, in one inexpensive extraction call that's part of every scan on every plan.
See it on your own domain — first dossier in about a minute.