Skip to content

Tool or agency

AI visibility tool, or an agency?

A tool tells you how assistants describe you. It does not change the answer. Knowing which half of the problem you have decides the purchase.

The short answer

Buy a tool if you can act on what it shows and only lack visibility into the problem. Buy an agency or a consultant if the diagnosis is clear and nobody has the hours or skill to rewrite pages and correct third-party sources. The common expensive mistake is paying agency rates for what is effectively a dashboard subscription with a monthly call.

What ranked on 23 September 2026: Searches in this space return tool vendors, tool listicles written by tool vendors, and agencies. Almost nothing compares the two categories against each other, because almost everyone publishing sells one of them.

Two different purchases

They are frequently sold as alternatives and are not. One is measurement, the other is labour.

  • A monitoring tool samples prompts on a schedule and reports mentions, position in the answer and often the sources cited
  • An agency should be doing editing, structural work and third-party correction, which no tool performs
  • Either can be run by one competent in-house person, which is the cheapest arrangement and the least sold
  • A tool has a floor price and predictable cost; agency pricing in this category is published almost nowhere
  • The tool's output is only useful if someone acts on it within the month, otherwise you are paying to watch

Things that sound right and are not

  • ClaimThe tool includes recommendations, so we do not need help.

    Recommendations are generic by construction. They can tell you a page is not extractable; they cannot rewrite it in your voice with your claims.

  • ClaimThe agency includes monitoring, so we do not need a tool.

    Fine, until you change agency and discover the baseline lived in their account. Own the measurement regardless of who does the work.

  • ClaimWe will do both to be safe.

    That is the most expensive arrangement and the one sold hardest. Start with measurement, prove where the gap is, then buy labour against a specific gap.

Where to start

  1. 01Sample your own prompt set manually for two weeks. It is free and it tells you which half of the problem you have.
  2. 02If the answers are wrong about facts, the work is editing and third-party correction. If they are right but never mention you, it is authority and consensus, which takes longer.
  3. 03Buy the tool when manual sampling stops being practical, and keep the account in your own name.
  4. 04Buy labour against a written scope with execution included, not against a monthly report.
  5. 05Re-run the same prompt set every month, whatever you buy, so you can tell a model update from your own work.

Questions

Are AI visibility tools accurate?
They are sampling a non-deterministic system, so treat their numbers as survey results rather than counts. Two tools can report different figures for the same week and both be defensible, which is why keeping one fixed method matters more than picking the best vendor.
What should a tool cost compared with an agency?
Tools have published subscription pricing; agency work in this category rarely does. The useful comparison is not price against price, it is what changes on your site in month one under each.
Can we start with neither?
Yes. A spreadsheet and a fixed prompt list run monthly is a legitimate baseline and costs an hour. Most organisations should do this before buying anything.

Related

Doing this yourself and want a second opinion on what the assistants say about you? Tell me what you are seeing, or read the AEO Playbook for the whole framework.