FIND OUT WHY THE AI RECOMMENDS SOMEONE ELSE.

Captus Proof watches what ChatGPT, Perplexity and Gemini actually say when a customer asks for a recommendation in your category: then hands you the prompt, the answer and the citation behind every number, instead of a score you're asked to trust.

Book a demo See how it works

AI engines recommend your competitors. Captus Proof shows why.

Every other tool hands you a black-box visibility score. Captus Proof queries three engines live and attaches the raw prompt, transcript, and citations to every number.

Before Captus Proof

Guessing why ChatGPT mentions competitors, testing prompts manually across personal accounts, and black-box visibility scores with no sample size or evidence.

After Captus Proof

Three engines queried live through their own APIs. Share of voice rolls up with statistical confidence, prompt transcripts, and every cited competitor domain attached.

One real question, asked the way a customer would ask it.

Not a keyword volume estimate. The question someone actually types, sent to each engine and read for what it says about you.

01

A real prompt

You write the question a customer asks, in their words. Proof keeps it as the unit of measurement.

"best all-purpose cleaner for a UK kitchen?"
02

Asked, not guessed

Sent live to each engine through its own API, repeated enough times to see past the variance.

chatgpt · perplexity · gemini
03

Read, not summarised away

Each response is scanned for your brand, the competitors you named, sentiment and every citation.

mentioned? cited? sentiment?
04

Disclosed, not scored

Rolled up into share of voice with the sample size and confidence label still attached to it.

36.8% · high confidence · n=38

It doesn't stop at the number.

The same prompt keeps going: the gap it exposes becomes a costed plan, and every plan item becomes the actual page, drafted and ready to check.

Every gap becomes one specific tactic, with a price on it

Not the same generic suggestion repeated for every prompt. A direct-answer page where a named competitor is winning. Structured content where no citation exists at all. Earned PR where third parties own the whole answer. Each with a planning-level cost band, one-off separated from ongoing, so you can see which part is a day of work and which is a retainer.

aeotracker plan generate --brand elbow-grease

One-off work
£1,300 to £3,200
Ongoing, monthly
£800 to £2,300
Direct-answer page for "Is this spray any good?" £150 to £400

Chosen because a named competitor currently owns this answer, not because it's the default tactic.

It writes the page. It doesn't invent the facts.

Title, meta description, body and FAQ schema, drafted from a real tracked gap. One hard rule: where a specific product fact belongs, the draft marks it and waits for a human, rather than inventing a plausible-sounding certification or test result. A generator that fabricates product claims is a liability, not a feature.

aeotracker content draft --brand elbow-grease

What's the best all-purpose cleaner for a UK kitchen? schema

Short answer: Elbow Grease is built with exactly this question in mind. [ADD: the product fact that answers this]

Why it matters here: how Elbow Grease compares against amazon.co.uk and wikipedia.org [ADD: honest comparison point]

Then it checks whether the fix actually worked

Mark a suggestion implemented and the history splits at that date. The dashboard compares the same prompts, on the same engines, before and after, and will tell you there isn't enough data yet rather than claiming a win it can't support.

aeotracker suggestions effectiveness

Schema markup added, 14 Aug proven

Share of voice on affected prompts: 21.4% before, 36.8% after. 38 executions either side.

Direct-answer page, 2 Sep not enough data yet

9 executions since implementation. Confidence stays low until the sample supports a claim.

What every other AEO tool leaves out

Compiled from independent reviews rather than vendor marketing. Everyone tracks mentions. The differences start immediately after that.

Capability Searchable Profound Otterly.ai Peec AI AthenaHQ Scrunch AI Captus Proof
Multi-engine mention tracking Yes Yes Yes Yes Yes Yes Yes
Shows the raw prompt, response and citation behind a number No No No No No No Yes
Discloses sample size and confidence alongside visibility figures No No No No No No Yes
Identifies which third-party domains own the winning citations No Yes No No Enterprise only No Yes
Estimates the cost of each content fix before you build it No No No No No No Yes
Drafts the replacement content Yes No No No No No Yes
…without inventing unverified product facts Unclear No No No No No Yes
Checks whether AI crawlers can reach your site No No No No No No Yes
Flat pricing, no per-engine credit system Unclear No Yes No No No Yes

"Content-fix suggestions" is marked yes wherever a vendor ships the feature at all, including where reviewers describe the implementation as thin or gated to a higher tier. Where a claim rests only on vendor marketing with no independent review, it's marked unclear rather than assumed either way.

Sources
  1. Otterly.ai reviews on G2: reviewers report a lack of actionable recommendations and historical data too shallow for trend analysis.
  2. SparkToro research on AI inconsistency: under 1-in-100 odds that ChatGPT returns the same brand list twice for the same prompt.
  3. Profound reviews on G2: broad platform coverage alongside a real configuration learning curve.
  4. Peec AI reviews via tryprofound.com and Surferstack: monitoring only at roughly $500/mo, with no gap analysis or writing tools, and an effective price 30-50% above the listed base once engines and countries are added.
  5. Scrunch AI review via Trakkr: improvement suggestions described as minimal, with a per-engine credit system that exhausts agency plans quickly.
  6. AthenaHQ review via Trakkr: credit-based pricing makes monthly cost unpredictable, and the citation engine is enterprise-only.
  7. Searchable product marketing: describes an agent that auto-generates AEO content. No independent review was found examining whether that content avoids inventing unverified product claims.

See your own category, with the receipts.

Give us your brand, three competitors and five questions your customers actually ask. We'll run them across all three engines and walk you through what came back: the numbers and the transcripts underneath them.

Book a demo Read the comparison

What a demo covers

  • Live category query

    Your brand and competitors queried live across ChatGPT, Perplexity and Gemini.

  • Full citation and prompt transcripts

    Walk through the raw engine output and the exact domains cited.

  • Costed content gap analysis

    Understand which gaps to fix first with disclosed price bands.

  • Direct Q&A with product engineers

    No sales scripts: ask about sampling methodology, APIs and confidence intervals.