Diagnose

Diagnose takes one question a buyer might ask, samples the assistants that can answer it, pulls in every signal Trakkr already holds about that question, and writes one brief explaining why the answer looks the way it does. It is the single-question read: Prompts watches a chosen list every day, Research sweeps the whole category once to find questions worth watching, and Diagnose is where you go when one of those questions deserves a real explanation with the evidence attached.

Open it when a position surprises you, when someone asks why you do not show up for something, before you commission a page you would rather justify first, or a few weeks after you shipped a fix. It is deliberate and capped, so it is not a page to use on every prompt you track.

Running a diagnosis

Type a question of three characters or more into the New diagnosis band and press Diagnose. The counter above it shows what is left of this month's allowance (for example, "21 of 25 diagnoses left"). Under the field, Trakkr suggests questions: ones whose latest diagnosis fell since the diagnosis before, or a few starters. Picking one only fills the field, so you can edit it before you start. While a scan is in flight the page shows each AI as it answers, so you see the finding land rather than a spinner. You can leave while it works.

Two things happen at once. The AIs are sampled, and every other signal Trakkr holds on this question is gathered into evidence packs.

The sampling asks the question the way a person would, and only in the modes each AI can honestly be asked in. Trakkr calls Claude and Gemini with no search tool, so they can only answer from training. Perplexity always searches. ChatGPT can do both.

AI	Sampled with search	Sampled from memory
ChatGPT	Yes	Yes
Perplexity	Yes	No
Claude	No	Yes
Gemini	No	Yes

The packs gathered alongside it are your tracked visibility score for this query, your ranking trajectory over recent weeks, phrasing variants of the question, the pages cited while the AIs answered, your own pages on the topic and the schema they carry, brand perception themes, Reddit threads that shape answers in your category, and any earlier diagnosis of the same question.

All of it goes into a synthesis step that writes the brief against a strict schema. The brief says what it was built from, and where an evidence pack was thin it says so rather than inventing detail.

Reading the brief

A finished diagnosis has four tabs: Brief, Answers, Pages, and Scans (once there is more than one scan of the question). The Brief reads top to bottom as one argument: a verdict, the answers from memory and with search beside your score, the answer set against its source, who each AI named, the pages the answers cite, and the fixes.

The verdict

A one-sentence verdict, with the scan's date, market, and answer count beside it. Under it, the From memory and With search bands each say how many of the sampled answers put you first, with a square per answer. Next to them, Score on this question gives your tracked score (see Tracked visibility below).

Answer vs source sets one AI's answer order beside the order of the cited page it leans on, and says whether the page lists you. If the page lists you and the answer does not, the problem is downstream of the page. If neither lists you, the pages the AI used do not include you.

At the foot, the brief says what it was built from: the answers sampled, the pages cited in this scan or in your tracked answers, and the search queries from that tracking.

Tracked visibility

Two places show a visibility score for a question. Both are a diagnosis's own score: the tracked score saved with that diagnosis, not a fresh recomputation.

In a brief, Score on this question shows the tracked score saved with this scan, its change since the previous scan of the same question, and a line over the earlier scans. Open in Prompts takes you to the question there. If the question is not tracked, the band says this diagnosis has no tracked score for the question (the brand's average answers a different question, so it is not shown) and offers Track this question. An old diagnosis keeps the score it saved, not today's.

On the entry page, Earlier diagnoses lists the latest diagnosis of each question: the question, the date, how many Scans it has, a Trend line, the diagnosis's Score, and Since last (the change from the diagnosis before it, or "first scan"). Narrow the list with Ranking, Not ranking, Improving, or Declining, or search the question text. Select a row to open that brief. A score in this table can differ from the question's tracked score on Prompts today, because Prompts moves every day and the table keeps what the diagnosis saved.

The pages that write this answer

Pages AI answers cite lists the pages cited, with how many answers cited each one, the order the page lists brands in, and whether the page names you, is yours, or leaves you out. Open a row for the receipt: what the page actually says, read from the page. The Pages tab shows every cited page; the Brief shows the first few.

A source you never want counted can be ignored from its row menu (Ignore this source), which applies to this brief and every later one. Ignored domains are listed on the Pages tab under Ignored sources, each with a Restore button.

Who wins this answer

Named first shows which brand led the sampled answers. Open it in Competitors to compare the brands. Every AI shows each AI and mode with where you came (#1 to #6, or no place), how often you were first, and how many pages it cited. A dash for an AI in a mode means it cannot be asked that way, not that it never named you.

Who names you

Read Every AI in two modes: With search and From memory. The combinations need different work.

Open the Answers tab to read the raw text of every sampled answer, one AI at a time.

The fix plan

The Brief ends with up to three fixes, the most serious gap first. Fix 1 · Start here is the one to ship first. Each fix carries what to do, why, an owner and timeframe, an estimate of how many of the sampled answers it could reach, and a first step. The reach is an estimate, never a measured change.

Owners are functions, not people: Content, Marketing, Dev, Product, or Partnerships. Ask Agent opens a conversation about the question. Send fixes to Actions puts the fixes in your queue.

Acting on a diagnosis

Hand the fixes to Actions. Select Send N fixes to Actions to move them into your Actions queue as in-progress work with the diagnosis evidence attached. Once every fix is there, the button reads Open in Actions.

Track the question. Track this question beside the score adds this exact question to your active Prompts list, so it joins the tracked scan and the score starts filling. If it is already tracked, the page says "Already in Prompts." Tracking does not duplicate the fixes, which use the normal Actions path.

Scan again after you ship. Scan again starts a fresh diagnosis of the same question and uses one diagnosis from your allowance. The earlier result stays in Scans. On the Scans tab, What changed since the last scan opens the difference. A later scan can show that the answer moved. It does not claim your work caused the move, which is what Results is for.

Export prints the diagnosis in place (The diagnosis, PDF) or downloads the cited pages (Sources, CSV). Share copies the link (for teammates only) or the verdict as text.

Limits

Diagnoses are counted per brand, per calendar month, and the count resets on the 1st. The counter above the field shows where you are. When you reach the cap, the band says so and points you to plans; every earlier diagnosis stays open.

Enterprise follows the Scale allowance. A Diagnose link from elsewhere in Trakkr can land on an existing brief rather than spending a new scan: if the same question was diagnosed for the same market in the last 30 days, the page opens that one. Typing a question into the field yourself always starts a new scan, as does Scan again.

Common questions