Skip to main content
LLM Metrix
Competitive · Benchmarking

Beat your competitors
in every AI answer.

Track the competitors you name, on the prompts you scan, across every major LLM. AI share of voice, mentions and answer position side by side. See where you lead, where you trail, and which prompts to act on.

Start freeBook a demo

Free forever plan. No card required

app.llmmetrix.com/dashboard/competitors
Illustrative, sample figures in the product's real layout
Share of voice · All engines

Of the brands this engine named across your tracked prompts, how much of the conversation each one owns.

#1
Your brand
27%
#2
Acme.io
26%
#3
Northwind
22%
#4
Globex
15%
#5
Initech
9%
Gap to top rival
+1 pts
Mentions counted
325

Works with the AI engines your customers use

ChatGPTPerplexityGeminiClaudeGrokMeta AIDeepSeekGoogle AI OverviewsMicrosoft Copilot

Why it matters

Your share of voice, next to theirs.

Track your rivals

Add direct, indirect and aspirational competitors and benchmark on every prompt you care about.

Head-to-head leaderboard

See how you and every competitor stack up on mentions, share of voice, and where they placed inside each answer.

Scan a competitor as its own subject

Run your prompt set against a competitor and get their visibility score, mention rate, share of voice and average position ranked beside yours in one table.

Head-to-head tracking

See how much of each engine's conversation is yours versus a given competitor over time, and get alerted when a competitor gains 10 points or more of share of voice against your previous scan.

Share of voice tracking

Your share of the mentions in the competitive pool, plus each competitor’s own share plotted scan by scan. Older scans show a gap rather than a fabricated zero.

Category context

Benchmark yourself against every competitor you've added at once, for a category-level view rather than one-on-one comparisons.

Mechanics

From rival names to head-to-head numbers.

  1. 01

    Name your rivals

    Add each competitor as a name on the project. From the next scan onward, every delivered answer is checked for those names, so rivals are measured on your prompts rather than scraped from their marketing.

  2. 02

    Every engine gets identical inputs

    All active prompts go to every enabled engine in one scan, and one credit covers a prompt across all of them. Errored answers drop out of every downstream figure instead of counting as losses.

  3. 03

    Each scan freezes its own numbers

    When a scan completes, its mention shares, pooled, per engine and per rival, write onto the scan row and stay immutable, so history reads lean rows and cannot be quietly recomputed later.

  4. 04

    Placement is kept per mention

    Where a rival appeared inside an answer is stored as a one-based place: first-named is one, a mention with no readable order stays unranked. Typical placement then separates leaders from fine print.

  5. 05

    Movement raises the flag

    Every scan is differenced against the one before it. A rival gaining 10 points of share of voice in one step, the threshold is editable, fires the competitor-spike alert beside your own movements.

Who it's for

Who benchmarks, and why

Agency lead reporting against named rivalsClient decks need the rival column, not just your brand’s line.

You retain clients in categories with obvious named competitors and owe each one a recurring competitive read.

  • Head-to-head tables put your client beside every tracked rival on identical prompts, ready to lift into the deck.
  • Token-gated report links reach the client without a seat, and the print view renders the scan as a branded document.
  • One workspace per retainer keeps each client’s rivals, prompts and history separate, within the plan’s workspace cap.
  • White-label branding carries your agency’s logo and theme across dashboard, reports and email chrome.
  1. 01Add each client’s tracked rivals on their project’s Competitors page.
  2. 02Run the first scan and check the benchmark table ranks everyone together.
  3. 03Send each client a token-gated report link from the Reports page.

Metrics this role tracks: Share of voice · Visibility score · Mention rate · Position drift

Open Reports
PM positioning against the category leaderYou need what engines actually tell buyers about the leader versus you.

You own positioning in a category with one dominant incumbent and want recorded evidence rather than sales anecdotes.

  • Read the leader’s answers next to yours, verbatim per engine, full text, no paraphrase, no summary.
  • Typical placement shows whether the leader’s mentions land in the first sentence or the fine print.
  • Per-engine share of voice for the scan you open shows which engines favour whom today.
  • Battlecards built on stored answers survive the hearsay objection, because every line traces to a scan.
  1. 01Scan the category leader as its own subject from the Competitors page.
  2. 02Read the leader’s answers verbatim beside yours, per engine.
  3. 03Check typical placement to see whether their mentions lead or sit in the fine print.

Metrics this role tracks: Share of voice · Position drift · Mention rate · Visibility score

Open the Competitors page
Sales enablement lead arming reps with category positioningReps need the rival column of the story, not another feature slide.

You build the battlecards and objection handling for a team selling against the same named rivals every quarter.

  • Lift the head-to-head table straight into the battlecard, your brand beside each named rival, measured on identical prompts.
  • Quote engines verbatim per engine instead of a rep’s paraphrase, because every line traces to a stored answer.
  • Print the scan as a branded document with answers side by side for the leave-behind.
  • Export the underlying scan to CSV or JSON when the enablement stack lives in spreadsheets.
  1. 01Add the three rivals reps hear most on the Competitors page.
  2. 02Run a scan and lift the head-to-head rows into the battlecard.
  3. 03Print the latest scan with answers side by side for the field.

Metrics this role tracks: Share of voice · Position drift · Mention rate · Visibility score

Open Reports
Founder entering a crowded categoryPick the few brands your buyer actually compares, then watch the split move.

You have launched where several incumbents already get named, and you need category proof without an enterprise budget.

  • Track the shortlist behind your buyer’s decision, up to 20 rivals per project, on every plan including Free.
  • Parity is simple arithmetic: against two tracked rivals a third of the mentions is even; against nine, ten percent is.
  • Each rival’s share plots as its own line over time, so a squeeze builds visibly before it costs you the lead.
  • A newly added rival starts a short line, earlier dates show a gap rather than a flat invented zero.
  1. 01Add the shortlist of brands your buyer compares on the Competitors page.
  2. 02Watch each rival’s share line build over the first few scans.
  3. 03Turn on the competitor-surge alert in Settings → Notifications so a squeeze reaches you early.

Metrics this role tracks: Share of voice · Competitor surge points · Mention rate · Visibility score

Analyst tracking rival momentum month over monthYou report direction and magnitude, and you have to be able to reconstruct both.

You maintain a competitive tracker and need defensible month-over-month deltas rather than impressions.

  • Per-rival share-of-voice series come from values frozen at scan time, so prior periods cannot be silently recomputed.
  • The benchmark table carries each subject’s previous score beside the latest, making momentum a column rather than a memory exercise.
  • Set the surge threshold to your reporting granularity, the default flags a rival gaining 10 share points between scans.
  • Open the scan behind any point and export it to CSV or JSON when the spreadsheet is the deliverable.
  1. 01Add the rivals on your tracker list to the project.
  2. 02Read each subject’s previous score beside the latest in the benchmark table.
  3. 03Export the scan behind any notable point to CSV or JSON for the deck.

Metrics this role tracks: Score change · Share of voice · Competitor surge points · Position drift

Foundations

What is AI share of voice?

Share of voice is your slice of the mentions in a competitive set: your mentions divided by everyone’s, you plus the rivals you track. Mentions are the unit rather than positions because a generated answer has no ranked list to take a position from; an answer either names you or it does not, and where in the answer you sit is a separate question. The figure is computed once, from the answers that actually came back, when each scan completes.

Worked example: across a scan, engines name your brand 25 times, Competitor A 18 times and Competitor B 15 times. Your share is 25 ÷ 58, about 43 percent. The denominator counts only the competitors configured on the project; any other brand an answer happens to name is discarded before the ratio forms. That is deliberate, counting every name makes the number a measure of answer length, and it is also why this figure reads higher than a category-wide share from another tool.

Keep share of voice distinct from mention rate. Named in 25 of 100 delivered answers is a 25 percent mention rate, presence. Forty-three percent share is competitiveness. Both plot side by side on the dashboard because they answer different questions and can move in opposite directions. Parity also scales with your set: against two tracked rivals a third of mentions is even footing, while against nine rivals ten percent is. Read any percentage band against the size of the list it divides by.

  • Adding a rival lowers your share as arithmetic; removing one flatters it. Treat the list as part of the measuring instrument and avoid reshuffling it mid-quarter.
  • A scan where nobody was named has no share at all, it plots as a gap rather than as a confident zero percent.
  • A rival added today starts its line where tracking started; earlier dates stay empty instead of inventing history.

Method

How do you benchmark competitors in AI answers?

Start with names. Each competitor is a name on the project, up to twenty per project on every plan, and the analyzer checks every delivered answer for those names. Nothing contacts the rival: no crawl of their site, nothing in their server logs. The comparison is fair by construction, because your prompts go to the same engines at the same time on both sides of the table.

A rival can also be scanned as a subject in its own right. Your prompt set runs against their brand and produces their own visibility score, mention rate, share of voice and average position, held exactly the way yours is. The benchmark table takes the latest completed scan per subject in your home market and ranks everyone together, carrying each subject’s previous score alongside the current one so momentum reads directly off the row.

Inside individual answers, placement is ordinal. Where each rival sat among the options an answer named is stored as a place, first-named is position one, and a mention with no readable order counts as named-but-unranked rather than a fabricated slot. That gives the head-to-head a quality dimension beyond raw counts: leading a category from the first sentence and leading it from the fine print are different positions.

  • Direct, indirect and aspirational rivals work the same way, the analyzer matches names, so sub-brands are simply separate entries, each counting toward the cap.
  • Every scan re-checks all tracked rivals alongside your brand, so their figures move on your clock, not theirs.
  • Foreign-market scans stay out of the home-market table; the regional comparison view owns those rows.

Reading

Why does your share differ from engine to engine?

The pooled number hides divergence. ChatGPT can favour you while Gemini favours a rival, and a stable overall share can conceal one engine sliding for weeks. Share of voice therefore breaks out per engine for any scan you open, how much of the naming on each engine is yours versus each tracked rival, computed from that scan’s own answers, with the same denominator as the headline figure above it.

Read engine slices with their sample sizes attached. A slice’s denominator is that engine’s delivered answers and the mentions inside them, so an engine that answered twice cannot carry the weight of one that answered thirty times, and a hundred percent share built on two answers is arithmetic, not dominance. Cells round independently, which means a table can legitimately sum to ninety-nine or one-hundred-and-one.

Over time, your own share plots as one line per engine, from values frozen onto each scan, that is how ChatGPT diverging from Gemini becomes visible without reopening every scan. A competitor’s per-engine breakdown, by contrast, belongs to the scan you open rather than continuing as a series.

  • The pooled row reproduces the frozen headline share, so the matrix and the trend chart cannot disagree.
  • Engines sort strongest-first; one with no measurable share sorts last rather than reading as a genuine zero.

FAQ

Benchmarking questions, answered.

What is AI competitor benchmarking?
Competitor benchmarking means putting the same prompts to the same AI engines and measuring who gets named. Your brand and each competitor you track are scored from completed scans, visibility score, mention rate, share of voice and average position, so the comparison reflects what the engines actually said rather than estimates.
How is AI share of voice calculated?
Share of voice is your slice of the mentions in your competitive pool: your brand's mentions divided by all mentions of you plus every competitor you track. It is computed when each scan completes and plotted scan by scan, so your category position reads as one line you can put in a deck.
How many competitors can I track?
Up to 20 per project, on every plan including Free, it is a property of a project rather than a tier allowance. Each competitor is tracked across every engine and prompt in that project.
Can I track a competitor that has multiple brands?
Add each brand you want counted as its own competitor name, they are tracked independently and each one counts toward the 20 per project. There is no parent-company grouping that rolls sub-brands into a single line.
Will my competitors know I'm tracking them?
No. Tracking a competitor means asking the AI engines your own prompts and recording who they name, we never contact the competitor's site, so there is nothing in their analytics or server logs to see.
How often does competitor data refresh?
On your project's refresh cadence: automated scans run daily on paid plans and weekly on free, and you can scan yourself at any time. Every scan re-checks all of your tracked competitors alongside your brand, so their figures move on the same clock as yours.
Do competitor trend lines cover the time before I added a rival?
No. A competitor's share of voice is plotted from the scans that ran after you added them, so a newly added rival starts with a short line, earlier dates show a gap rather than a zero invented to fill the chart.
Is this different from the answer-ranking view?
Yes. Ranking reads one subject at a time, where a brand's mentions land inside each answer, first mention through absent. Benchmarking places every rival you track beside you in one table: mention counts, share of voice, and where each of them sat inside the same answers. The ranking view tells you where you sit; the benchmark tells you who is winning the category.
How do I start benchmarking a competitor?
Add each competitor's name to your project, up to 20 per project, and run a scan. The analyzer records every answer that names them, and the Competitors page fills in with the head-to-head table and trend lines once that scan completes. Nothing is installed on the competitor's site and there is no verification step, because tracking only reads what the engines say.

Keep exploring

Related features

Answer Engine Ranking

See exactly where your brand appears: first mention, prominent recommendation, or buried in fine print.

Multi-Engine Monitoring

Track your brand across ChatGPT, Perplexity, Gemini, Claude, Grok, Meta AI and DeepSeek: the seven major AI models in one view, plus Google AI Overviews, Google AI Mode and Microsoft Copilot on a weekly cadence for paid plans.

Visibility Score

A single composited score from mention rate, answer position and sentiment that shows how AI engines understand and recommend your brand.

In depth

New: Competitor Benchmarking Is Now on Every Plan

Competitor tracking is no longer gated by tier. Seeing who the engines name alongside you falls out of the scan you already ran. Here's what that gets you free, what a head-to-head benchmark costs in credits, and how to read both.

7 min read

The AI Visibility Metrics That Actually Matter

Mention rate, citation rate, share of voice, share of model. AEO has its own vocabulary of metrics. Here's what each one measures, when to use it, and how they fit together.

8 min read

Why Most AI Visibility Audits Never Change Anything

The audit finds forty problems, everyone agrees they're real, and six months later the visibility score hasn't moved. The failure isn't the analysis: it's that a findings list and a sequenced backlog are different artifacts. Fix what you control before chasing citations, and give every item an owner.

8 min read
Start on the free plan. No card, no time limit.

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.

Start freeBook a demo