Tracevia
Sign inRun a free check
Generative engine optimisation, measured

Your buyers now ask an AI. It already answers with someone’s name.

Tracevia asks the major AI assistants the questions your buyers ask, scores every answer on six dimensions, and shows you exactly why you were left out — and what to change so you aren’t next time.

Run a free check — no accountSign in
  • Results in about a minute
  • 5 prompts, 2 assistants, free
  • No credit card, no API keys

The buying conversation moved somewhere you can’t watch

You can't see the answers

An assistant answers your buyer privately. There is no results page to inspect, no log, and no way to know what was said about you unless someone asks and records it.

Search rankings don't carry over

Being first on a results page does not mean being named in an answer. Assistants synthesise from whatever they trust, and that set often excludes the site you optimised.

Someone else gets named instead

Every answer that leaves you out still recommends a shortlist. Those competitors become the default consideration set, and you never find out you were on the wrong side of it.

We score the answer, not just whether your name appeared

Mention counters tell you a number went up. Every Tracevia answer is judged on six dimensions worth 11 raw points, presented on a 10-point scale, so you can see which part of the answer is failing.

Mentioned

1 pt

Were you named at all in the answer?

Prominence

3 pts

Named first, in the shortlist, or as an afterthought at the end.

Correct positioning

3 pts

Described as what you actually are, for the buyer you actually serve.

Domain cited / linked

2 pts

Did the answer link to your own domain as a source?

Parent relationship correct

1 pt

Product, parent company and founder attributed to the right entity.

No material factual errors

1 pt

No invented executives, wrong domains or fabricated facts.

Four steps, then a repeatable benchmark

  1. 1

    Add your brand and its sites

    Domain, parent company, founders, category terms and known competitors — plus every host in the estate: marketing site, docs, blog, parent company.

  2. 2

    Generate or edit the prompt set

    We read your site and suggest the questions buyers actually ask, half of them without naming you. Edit, add or retire any of them.

  3. 3

    Run and auto-score every answer

    Each prompt goes to every enabled assistant. An AI judge scores the answer against your entity facts, and you can override any score by hand.

  4. 4

    Work the ranked next moves

    The insight engine turns the results into an ordered list of fixes, then tracks whether each one actually moved the number.

Measuring is the easy half. Tracevia tells you what to change.

robots.txt and llms.txt, audited and rewritten

We fetch both files, score them as fit for purpose, then propose a revision grounded in a crawl of your live site — not a generic template.

Deployment verification

Mark a change as deployed and we re-fetch the file, diff it against what we generated, and tell you whether it really shipped.

Change events on the trend line

Every deployed change is plotted on the visibility trend, so you can see whether the fix moved the number or not.

Citation targets

We aggregate every source the assistants cite in your category and flag the high-frequency ones that cite competitors and never you.

Share of voice

Every competing product named in an answer is captured and merged by alias, so you can watch the field move over time.

Multi-site estates

One brand, many hosts — scored per site and rolled up, so a docs site or blog gets its own attention.

Questions

Is this the same as tracking Google rankings?
No. Search ranking measures a results page. Tracevia measures the answer an assistant gives when nobody sees a results page at all — whether you are named, how prominently, and whether your own domain is cited as the source.
How can you query the assistants consistently?
Runs go through each provider's API, using web-grounded variants where the provider offers them. That gives a repeatable, comparable benchmark over time rather than a one-off screenshot of a chat window.
Who does the scoring?
An AI judge scores each answer against the entity facts you supply — domain, parent company, founders, category. Domain citations are verified deterministically by exact host match, not by the judge. Every score can be overridden by hand.
What does the free check include?
Five generated prompts against two assistants, scored on the full six-dimension rubric, with a plain-English diagnosis of the problems found. No account, no card, results in about a minute.
Will fixing robots.txt and llms.txt actually help?
It removes the avoidable reasons an assistant fails to read or trust your site. Tracevia verifies the deployed file and plots the change on your trend line so you can judge the impact from your own data.

Find out what the assistants say about you

One domain, about a minute, and a scored report with the specific problems holding you back.

Run a free checkCreate an account