First mention, or
fine print?
LLM Metrix parses every AI answer and tags your brand's position, first, prominent, mid-list or fine-print. Track it on your plan's cadence across every engine and prompt.
Free forever plan. No card required
Where LLM Metrix lands among the options each engine named.
“LLM Metrix is the most comprehensive platform for tracking brand visibility across AI engines.”
“Among the stronger options is LLM Metrix, which monitors mentions across major engines.”
“Other tools in this space include LLM Metrix and several smaller vendors.”
“The answer named four vendors and none of them was the tracked brand.”
Tracked competitors named in these answers, with their place inside the answer when the engine ranked them.
Works with the AI engines your customers use
ChatGPT
Perplexity
Gemini
Claude
Grok
Meta AI
DeepSeek
Google AI Overviews
Microsoft CopilotWhy it matters
Not just mentioned, placed.
Position-tagged answers
Every captured response is automatically tagged with your brand's rank in the answer.
First-mention rate
The share of delivered answers that name you ahead of everything else, as one figure per scan and a band on the position chart. Answers that never mention you count against it, so it reads as visibility rather than as a tally of your best results.
Position drift, on the chart
Plot your average position and the full first-to-absent mix across your scan history, so a slide out of the top bands is visible on the trend, and get alerted when your average slips a position or your first-mention share drops.
Per-engine ranking
See your average position engine by engine, with the mention rate and answer count beside it so an excellent average built on two answers is obvious, the individual answers are listed on the same page.
Sponsored blocks flagged
Answers carrying paid or sponsored placements, ad labels, shopping or product cards, are flagged on the Rankings page, so a bought slot never reads as earned position.
Recommendations from the same scan
The recommendations are generated from the scan that produced these positions, the engines you are weakest on and the prompts you were missed on entirely.
Position-tagged answers
Every captured response is automatically tagged with your brand's rank in the answer.
First-mention rate
The share of delivered answers that name you ahead of everything else, as one figure per scan and a band on the position chart. Answers that never mention you count against it, so it reads as visibility rather than as a tally of your best results.
Position drift, on the chart
Plot your average position and the full first-to-absent mix across your scan history, so a slide out of the top bands is visible on the trend, and get alerted when your average slips a position or your first-mention share drops.
Per-engine ranking
See your average position engine by engine, with the mention rate and answer count beside it so an excellent average built on two answers is obvious, the individual answers are listed on the same page.
Sponsored blocks flagged
Answers carrying paid or sponsored placements, ad labels, shopping or product cards, are flagged on the Rankings page, so a bought slot never reads as earned position.
Recommendations from the same scan
The recommendations are generated from the scan that produced these positions, the engines you are weakest on and the prompts you were missed on entirely.
Mechanics
From answer to position, per scan.
- 01
The prompt goes to every engine
Each scan asks every enabled engine the same prompt. Positions come from the answers that actually returned, errored calls drop out of every bucket rather than counting as absences.
- 02
The analyzer returns one ordinal
An analyzer reads each full answer and returns where your brand sat among every option that answer named, a one-based place, or null when the engine named you without ranking anything.
- 03
Rivals get places too
Every tracked competitor the answer names carries its own recorded place, 1 for the first slot and counting down, 0 when nobody was ranked, stored beside yours on the same answer row.
- 04
The scan freezes its mix
The five tiers are counted per scan alongside average rank and the first-mention share, written onto the scan row, so trends read frozen figures instead of recomputing history.
- 05
Demotions reach you
Average-rank moves of a position or more, ten-point drops in first-mention share, and per-row band crossings fire against the previous scan behind the Position change switch, reported per engine.
Who it's for
Who tracks their place in the answer
In-house SEO lead running rank-tracking for AI answersClassic rank tracking, except the results page is a paragraph.
You already run rank-tracking workflows and need the same discipline for answers that never show a numbered list.
- The Query Rankings panel lists every prompt with its per-engine chip, First mention through Absent, plus movement chips against the previous scan.
- The Position over time chart separates average rank from the tier shares, so a slide stays visible even while your mention rate holds flat.
- The volatility card grades the window Flat through Storm, so a turbulent fortnight is a measurement rather than a mood.
- The latest scan exports to CSV or JSON when the deliverable is a spreadsheet.
- 01Run prompt discovery and track the prompts buyers actually ask.
- 02Read the Query Rankings panel per engine after the first scan.
- 03Switch on Position change alerts in Settings → Notifications.
Metrics this role tracks: Average position · First-mention rate · Mention rate
Open RankingsDemand-gen lead optimising bottom-funnel questionsPipeline rides on the prompts where buyers shortlist.
You care about the commercial questions where being named first correlates with pipeline, and need to find them and defend them.
- Import your top Search Console queries as tracked prompts in one click, they land inactive until you activate them, deduped against what you already track.
- The Search↔AI gap names queries you rank for in Google that no engine cites you for, bottom-funnel demand currently leaking elsewhere.
- First-mention rate tells you whether the shortlist opens with you or with a rival, per scan.
- 01Connect Google Search Console on the Search page.
- 02Import your bottom-funnel queries as tracked prompts and activate them.
- 03Run a scan and check the first-mention rate on the commercial questions.
Metrics this role tracks: First-mention rate · Average position · Mention rate
Open SearchCompetitive intelligence analyst tracking rival placementWho got named ahead of you, in which engine, on which prompt.
Your job is explaining share shifts, and "we lost the answer" needs evidence at the level of individual responses.
- Each tracked rival named in an answer carries its recorded place, so a competitor moving ahead of you inside one answer is readable rather than guessable.
- Per-rival share-of-voice trend lines show a squeeze building across scans instead of surfacing once it has landed.
- Scan a competitor as a subject and read their score against yours on identical prompts.
- 01Add the rivals named ahead of you on the Competitors page.
- 02Reopen an answer where a rival moved ahead and read both recorded places.
- 03Watch the per-rival share lines build across the next few scans.
Metrics this role tracks: Share of voice · Average position · Mention rate
Open the Competitors pageBrand manager guarding earned placementA bought slot should never read as earned position.
You distinguish organic recommendation from advertising inside AI answers, and need the distinction recorded rather than eyeballed.
- Answers carrying paid or sponsored blocks, ad labels, shopping or product cards, are flagged per engine and per answer on the Rankings page.
- Band demotion alerts name the engine and the prompts that slipped, so recovery work starts somewhere specific.
- The Answer Archive shows what changed in an answer between scans, additions and removals highlighted, claims included.
- 01Review the paid-placement flags per answer on the Rankings page.
- 02Turn on band-demotion alerts in Settings → Notifications.
- 03Diff a changed answer in the Answer Archive before you respond.
Metrics this role tracks: Paid-placement flags · Average position · First-mention rate
Open AlertsContent editor rewriting for the answers that buried youThree engines name you mid-list on the exact question buyers ask.
You write the pages and need the specific prompts and engines demoting you, plus a spec for what to change next.
- The Query Rankings panel names the prompts holding you in mid-list or fine print, per engine, so the rewrite targets the answers doing the burying.
- Recommendations generated from the same scan arrive tagged critical through low, each carrying the finding behind it and the engines it targets.
- Content-gap rows generate a content brief on the recommendation board, a spec for the rewrite rather than a hunch.
- The Answer Archive diffs an answer against the prior scan, so you can tell whether the rewording moved anything.
- 01Shortlist the prompts where you sit mid-list or worse in the Query Rankings panel.
- 02Generate the content brief from a matching row on the recommendation board.
- 03Ship the rewrite and compare positions on the next scan.
Metrics this role tracks: Average position · First-mention rate · Mention rate
Open RecommendationsThe taxonomy
How do you classify positions inside an AI answer?
An AI answer has no ranked list to read a position off, so position is reconstructed from structure. An analyzer reads the full response and returns one number: where your brand sits among every option that answer named, all of them, not only the competitors you happen to track. A fixed ordinal rule then buckets that number, and the rule never sees the text: 1 is First, 2–3 Prominent, 4–7 Mid-list, 8 or later Fine-print, a mention with nothing to order it Fine-print too, and no mention at all Absent.
Two consequences are worth knowing before you read your own rows. The tier is your brand’s alone, emphasis and framing were weighed earlier, by the analyzer producing the rank, not by the bucketing. And fine-print does double duty: it catches both the genuinely buried mention and the unordered one, which is why a scan of mostly-unranked mentions can freeze a flattering average rank. The count of mentions that carried no rank sits beside the average for exactly that reason.
- First mention = #1 · Prominent = #2–3 · Mid-list = #4–7 · Fine-print = 8th or later, or mentioned but unranked · Absent = not mentioned.
- Where each tracked competitor placed in the same answer is recorded too, one-based, or zero when the engine ranked nobody.
The metric
What is first-mention rate and why does it matter?
First-mention rate is the share of delivered answers that name your brand ahead of everything else. The denominator is every delivered answer, including the ones that never mentioned you, so it reads as visibility rather than as a tally of your best results. Because rank is defined among all named options, first place genuinely means nothing at all was named ahead of you.
It is the sharpest single number on the page because early mention sets the frame: the brand named first anchors the recommendation, and later names are read against it. When you want the other reading, of the answers that mentioned you at all, how many led, divide the first-place share by the share that was not absent.
Per engine, the Rankings page carries an Average Position table, average rank, mention rate and answer count side by side, because pooling hides the finding worth acting on: first on one engine and absent on another averages into a mediocre middle, and an excellent average built on two answers should look as thin as it is.
Drift
How do I catch position drift?
Watch the Position over time chart: one plot of your average rank, and one of the share of delivered answers sitting in each tier, oldest first. Shares rather than raw counts is the deliberate choice, buckets are counts of answers, so a project that adds tracked prompts would otherwise watch every band rise and read pure volume as improvement. A rising average, or the first band giving way to mid-list, is drift in progress.
A volatility card condenses the same series into a 0–100 index computed over consecutive-scan pairs, rank movement and mix turnover equally weighted, and bands it from Flat through Storm. It needs at least three scans before it speaks, because one pair is a blip rather than a trend.
Three alert forms sit behind the Position change switch in Settings → Notifications: your average rank moving a position or more between scans, first-mention share dropping ten or more points, and individual engine-and-prompt rows crossing a band downward, First → Prominent, Mid-list → Fine-print, named → Absent, reported per engine. Wording is a separate read: the Answer Archive diffs an answer against the prior scan with additions and removals highlighted.
FAQ
Ranking questions, answered.
How do you classify ranking positions?
Does ranking work without explicit lists?
Can I track ranking on long-tail prompts?
Where do my competitors rank in the same answer?
How often are positions updated?
How is this different from Google rank tracking?
Which engines can I track positions on?
Keep exploring
Related features
The AI era of search
is already here
See what AI engines say about your brand before your competitors do. Start free today. No card required.