Competitive Intelligence

AI Competitor Tracking

When a buyer asks an assistant who they should be looking at, it returns a short list of names. Competitor tracking tells you whose names those are, where yours sits among them, and which sources the model reached for to decide.

TL;DR

AI competitor tracking runs one fixed set of buyer questions against the AI engines on a schedule and records every brand named in the answers, not just yours. A rank tracker follows a rival's page. This follows whether the model recommends them instead of you.

4
signals read per rival
1
prompt set for everyone
discovered rivals, paid plans

Why this is a separate discipline

Competitive intelligence has always worked by finding things competitors published. A press release, a pricing change, a review, a ranking. Something exists at an address, a crawler finds it, it lands in a dashboard. That chain is the architecture of every competitive tool in use today.

An AI answer breaks the first link. When an assistant recommends three brands in your category, nothing is published: the answer is written for one person and then it is gone. Your competitor was recommended and you were not, and no crawler-based tool can tell you it happened, because there is no document to find.

So the only way to see it is to ask. Send the buyer questions yourself, on a schedule, and record what comes back. The measurement is the same for every brand named in the answer, which is what makes it comparative rather than a report about yourself.

Words you'll see, in plain English

Prompt set

The fixed list of buyer questions run on a schedule. Fixed wording is what makes week-over-week movement mean something.

Mention rate

How often an answer names a brand across that prompt set. Read for you and for every rival in the same run.

Position in answer

Whether a brand appears in the opening sentence or is the fourth name in a list. Not a page ranking.

Share of voice

Your presence relative to named competitors across the same questions and the same engines.

Discovered competitor

A brand the models name alongside you that you never entered. Tracked without limit on paid plans, and often the most useful row in the table.

Citation

When an engine links a domain as a source under its answer. Shows which rival's content the model is leaning on.

What you can actually see

Four readings, taken for every brand named in the answer rather than only for yours.

Who gets named instead of you

For every tracked question, the full list of brands the engine returned, in the order it returned them. The absence of your name is the finding.

Where each brand sits in the answer

Being the first name in a paragraph and being the fourth item in a list are different commercial outcomes. Position is tracked per brand, per engine.

Which sources back a rival up

The citations under an answer show which pages the model reached for. When a competitor's visibility climbs, the citation list usually explains why.

How the description differs by engine

A rival can dominate one assistant and be absent from another, because each has its own retrieval pipeline and training cutoff. Averages hide this.

How tracking works

Four steps, and the second one is where most competitive reporting goes wrong.

01

Add competitors, and let the models add more

Enter rival brand names and their aliases so variations are counted as one entity. The models also surface brands you did not enter, named alongside you in real answers, and on paid plans those discovered competitors are tracked without counting against the cap.

02

Run one fixed prompt set for everyone

Your configured questions run on a schedule, and every brand named in the returned answers is recorded from the same run. Same questions, same engines, same day, which is what makes the comparison fair.

03

Read the answers side by side

Visibility, position, sentiment and mention context, per competitor and per engine, rather than a single blended score that hides where you are actually losing.

04

Get told when something moves

Alerts when a competitor's visibility shifts meaningfully, plus scheduled reports summarising the competitive picture, so nobody has to remember to check.

What a finding looks like

An illustration of the shape these findings take, not a customer story.

Tracked prompt
"which vendors should I shortlist in this category?"
Week 1
  • 1. Rival A
  • 2. You
  • 3. Rival B
Week 6
  • 1. Rival A
  • 2. Rival C (you never entered this one)
  • 3. Rival B

The useful part is not the drop, it is the citation list. A brand you were not tracking took the slot, so the first question is which sources the engine started citing for it. If those citations point at an independent category roundup, publishing another blog post will not answer it. If they point at original research, a comparison page will not either. The response depends on what kind of source moved, which is why the citations matter more than the ranking.

What to do with a finding

Tracking is only worth the subscription if something happens after the alert.

01

Read the citations before anything else

A rival's jump almost always traces to a specific source the models started reaching for. Find that page first, because it tells you what kind of asset moved the needle.

02

Match the asset type, not the headline

If an independent roundup is doing the work, another blog post will not answer it. If original research is doing the work, a comparison page will not either.

03

Check the prompts where you are absent entirely

Being fourth is a ranking problem. Not appearing at all is usually an entity or coverage problem, and the two need different work.

04

Tag your own pushes so you can attribute

Mark the date of every launch, PR wave and content push. Without the tag you will watch a line move and have no idea which lever moved it.

See who the models name in your category

A free scan returns the answers verbatim, with every brand the engines named.

Run a free scan

How many competitors, and how often

Entered competitors are capped by plan. Discovered competitors, the brands models name alongside you without being asked, are unlimited on every paid plan. Enterprise is quoted on custom volumes rather than a fixed cap. Eight AI engines are tracked in total; how many your plan includes varies, and the rest are available as an add-on.

PlanCompetitorsTrackingAI engines (of 8)
Free5Weekly3
Starter10Weekly4
Growth15Daily5
Scale20Daily5
ProUnlimitedDaily5

Current limits, and they can change. See full plan details.

Setup checklist

  • Every rival is entered with its aliases, so name variations count as one brand
  • The prompt set is fixed in wording and version-controlled
  • Prompts are labelled by intent, with commercial ones weighted more heavily
  • Visibility is read per engine before it is averaged
  • Discovered competitors are reviewed, not just the ones you entered
  • Alerts go to someone who will act on them, not to a shared inbox
  • Every launch and campaign is tagged so movement can be attributed

Frequently asked questions

How can I monitor my competitors' performance in AI search?

By running one fixed set of buyer questions against the AI engines on a schedule and recording every brand named in the returned answers, not just your own. Because an AI answer is generated on request and never published as a page, there is nothing for a crawler to find. The questions have to be sent deliberately, and the wording has to stay fixed for week-over-week movement to mean anything.

How do we track competitor performance in Google AI Overviews?

The same way as any other surface, and separately from the rest. AI Overviews is tracked as its own engine because it has its own retrieval pipeline, so a competitor can hold the AI Overviews answer in your category while being absent from ChatGPT. Reading one blended number across engines hides exactly the gap you need to see.

Can I monitor competitor sentiment in AI responses in real time?

Not in real time, and it is worth being precise about that. Tracking runs weekly on the Free and Starter plans and daily from Growth upward. That is a deliberate cadence rather than a limitation: AI perception moves on retrieval and model-refresh cycles, so a brand's description rarely changes minute to minute, and reading it more often mostly adds noise.

How is this different from a competitor rank tracker?

A rank tracker records where a competitor's page sits in a list of search results. This records whether an AI engine names the competitor inside the answer it writes, where in that answer they appear, and which sources it cited to get there. Different surface, different unit: the page versus the brand.

How many competitors can I track?

Five on the Free plan, ten on Starter, fifteen on Growth, twenty on Scale, and unlimited on Pro. Enterprise is quoted on custom volumes. Separately, every paid plan tracks unlimited discovered competitors, meaning brands the models name alongside you that you never entered, which is often where the surprises are.

Do competitors know they are being tracked?

No. Nothing is sent to them and nothing touches their systems. The questions go to the AI engines, exactly as any buyer would ask them, and what is recorded is the public answer that comes back.

What if a competitor we did not enter keeps appearing?

That is a finding rather than a nuisance. Models name the brands their sources associate with your category, so a name you did not expect usually means either a genuine competitor you had not counted or a positioning overlap worth understanding. Every paid plan tracks discovered competitors without limit for that reason.

How often should the competitive picture be reviewed?

Monthly for the trend and weekly for the alerts. Mention rate and position move slowly enough that a monthly read is where the story is; alerts exist so a sharp move does not wait for the monthly meeting.

Related features

Limited: 7-Day Free Trial

Monitor Your Competitors Today

See which rivals AI engines recommend in your category, and which sources put them there.

No credit card requiredSetup in under 5 minCancel anytime