Skip to main content

Build log

How do teams get started with an AI visibility agency?

·

At Sandcastle Labs, teams get started with an AI visibility agency in three steps: a free scan, a 30-minute call, then a manifesto that opens the 90-day sprint. The scan is the proof of concept for the problem: it shows what assistants say about you before you sign anything. It does not prove we can fix it.

How do you get started with an AI visibility agency?

Here is the order with us, what each step asks of you, and what you have at the end of it.

StepWhat you doWhat you have afterCommitment
1. Free scanEnter your website and a work emailA report on what assistants say about youNone. No card
2. 30-minute callBring the report and tell us what is happening at your companyA plain answer on whether the sprint fitsNone
3. ManifestoSign, then work with us on who you are, whom you serve and how you want to be foundThe document that drives the content strategyThe 90-day sprint
4. Day 90Choose how to continueThe pages, the fixes and a second read of what movedStay monthly, or keep our engine and run it yourselves

You can start at step 2. If you already know the problem is real, book a call. All three of us see the request, one of us usually replies the same day, and we bring what assistants say about you today.

What does the free scan show?

The free scan checks 3 AI engines across up to 20 buyer questions and emails you the report. It shows which engines mention your company when buyers ask without using your name, which other companies appear in those answers, what AI gets wrong about you, and recommendations with a clear place to start.

Three practical details, so nothing surprises you:

  • You need a work email, and it is one free scan per email.
  • We send a few follow-ups over the next month. You can unsubscribe anytime.
  • There is no card and no call required to get the report.

If all you want is the technical fixes, the scan may be as far as you need to go: take the report and make them yourself. We won’t discount the sprint to make it fit.

Is a free scan a real proof of concept?

For the problem, yes. For the fix, no. A free scan gives you a dated read of your own company on questions your buyers ask, so you can stop guessing about whether assistants name you. It cannot show that any agency, including us, will change the result, because nothing has been changed yet.

It also has limits as a measurement. It covers 3 engines, while the dated read that opens the sprint covers 5. It is one read, and answers vary from one asking to the next: the sample reports on the scan page, measured in October 2026, carry margins of ±9 and ±10 at 95% confidence. Treat a small gap between two scores as noise.

Proof of the fix is a before and after on the same question set. Our published receipt is one: a developer-tools company went from 1.25% to 5.8% of category answers citing its own domain in two weeks, 7 of 120 answers in the July 2026 read. Its site team shipped linking work in the same window, so we call that strong evidence rather than proof. It is one company in one category, and it is the only receipt we have published so far.

What should you have ready before the call?

Nothing needs preparing beyond the report, if you have one. What decides fit is how your company works, and these are the things we will ask about:

  • Buyers use assistants. People ask AI assistants about what you sell. Software, e-commerce, real estate and businesses with several locations usually pass.
  • You can ship. A change to your website goes live within a week.
  • One person can sign. Buying does not run through procurement or an RFP.
  • There is a mandate to grow. Usually something has just changed, like a new market, a stalled deal or an assistant describing you wrongly.

Most of the companies we work with have roughly 20 to 300 people. If you are outside those lines, we will say so on the call. The work with us page has the longer version, and the pricing page has the flat price, which is the same whatever your size.

What should you ask any agency before you start?

These hold whether you talk to us or to someone else:

  • What is the dated read before the work, and will you repeat the same question set after it? Without both, nobody can tell you what moved.
  • What do I keep if I leave? With us, you can cancel anytime and keep everything built.
  • Show me something that did not move. A firm that only reports wins is leaving out part of the record.
  • Do you claim to control what an assistant says? We don’t. We change what it reads, and the answer follows from what is there to read.

What we’re still figuring out

  • How well a free scan predicts sprint fit. We have not published any numbers on that.
  • A second receipt, so the case for the fix does not rest on one company.

For agents: try this yourself

  • Read your own baseline. Ask an assistant five questions a buyer of [what we sell] would ask without naming any vendor, and record the date and whether [company] appears in each answer.
  • Check an agency’s proof. For [agency], find one published before and after result, and report its date, its question set and any confounders it names.

How this was made

Drafted by Claude from our live scan, pricing, receipts and work-with-us pages, then edited and published by Chase.

See how the Chronicler works →

Try this with your own agent

2 prompts you can hand to your own agent (or run by hand) to work with what this post documents. Edit the bracketed parts for your context.

Read your own baseline

List five questions a buyer of [what we sell] would ask an AI assistant without naming any vendor. Ask each one, then report today's date, whether [company] appears in each answer, and which other companies are named instead.

Check an agency’s proof

Find one published before and after result from [agency]. Report the date of each read, the question set used, the number of answers behind each figure, and any confounders it names. If any of those are missing, say which.

More in Studio Build