Skip to content

FAQ

Frequently asked questions

How we measure, what it costs, how an audit runs and what we do with your data. If your question is not here, email info@recommendable.co.

About Recommendable

What is Recommendable?

Recommendable helps brands get recommended by AI assistants. We measure whether ChatGPT, Claude, Gemini and Perplexity recommend your brand, and why or why not, together with your site and your market in Google. The founders read every finding themselves and deliver an advice report with an action plan sorted by priority. Recommendable is a product of Think again. B.V., an AI consultancy from Utrecht, the Netherlands.

Read more

Is Recommendable a prompt tracker?

No. A prompt tracker sends a fixed set of prompts every day and counts whether you are mentioned. Recommendable measures repeatedly, with the margin shown, whether you are recommended, including in whole buying conversations, and turns what it finds into an action plan. You run it as a periodic audit, with a re-measurement to see what really changed.

Read more

Does Recommendable write content?

No, on purpose. Content generated at the push of a button is mediocre, and models see through it. We tell you per page what is missing, why it matters and what it earns you. You hand that brief to your writer, your agency or your own AI assistant.

What don't you do?

We don't write content. You get a briefing per page that your writer, agency or own AI assistant can work from. We also don't promise a place in an AI answer: a model decides for itself, and the same question gives a different answer every time. We don't measure yet what people search for inside YouTube and TikTok. And we are not a daily tracker: we measure per audit, repeatedly and with the margin shown.

Which languages and markets does Recommendable cover?

In Dutch and English: the analyses, the advice report and the presentation. You choose the market per brand, and the questions we measure come from that language. We measure extra markets as separate brands, so each country gets its own numbers.

How we measure

Which AI assistants does Recommendable measure?

ChatGPT, Claude, Gemini and Perplexity, through the providers' APIs and never through logged-in accounts or by reading the chat app. For each assistant we choose, as far as possible, the model ordinary users get, and we pin that choice. When the model changes, you see it as a break in the line, and it is never silently averaged with older measurements.

Read more

Do you also look at Google, YouTube and TikTok?

Yes, as far as we can measure it. For Google, the AI Search Website Audit measures the technical SEO foundation of your site, and the market analysis measures search volume and your Share of Search. YouTube, TikTok and Reddit we read as sources: what is said about you there and when assistants cite it. What people search for inside YouTube and TikTok we do not measure yet.

Read more

Do you measure Google AI Overviews?

Yes, per topic. We read the Google results for a topic's search terms and record whether an AI overview appears, which sites it cites and whether yours is among them. AI Mode is measured separately, because it cites other sites. Every reading is kept, so you see how often you appear over the weeks. It is a citation measurement per search term; the repeated measurement with a margin is what we do for the four assistants.

Why is there a margin on every number?

A language model answers the same question differently every time, so one measurement says little. We ask every question repeatedly and show next to every number the 95% band (Wilson interval) in which the real value lies. The label Robust, Adequate or Minimal tells you how much weight to give it, and the label Interpretation marks an interpretation that is not itself a measurement.

Read more

What happens when a measurement fails or returns too few answers?

A call that fails because of a rate limit or an outage does not count as 'not mentioned'. It leaves the denominator and is reported separately. Below five usable answers we make no statement, and the topic is shown as not measured.

What is the difference between memory and live?

In memory mode an assistant answers from its own knowledge, in live mode it may search the web first. We measure the two separately and show per topic in what share of the live answers a search actually happened. Perplexity always searches, so there we measure live only.

Do you promise more customers or revenue?

No. We measure whether you are mentioned, recommended and cited, and nobody in this market has proof that a ten-point rise brings a set number of customers. So we promise no ROI and put no amounts under a visibility figure.

Packages and pricing

What do the Recommendable packages cost?

The AI Search Website Audit costs 495 euro, the AI Search Strategy 1,495 euro and Strategy & Workshop 2,495 euro. Every package includes one re-measurement; after that, a re-measurement costs 195 euro after a AI Search Website Audit and 495 euro after a Strategy. All prices are one-off and exclude VAT, and you receive an invoice before we start.

Read more

What is the difference between the AI Search Website Audit and the AI Search Strategy?

The AI Search Website Audit (495 euro) reads your own site page by page: technical health, content, positioning and agent readiness. The AI Search Strategy (1,495 euro) includes the AI Search Website Audit and adds everything outside your site and inside the assistants: your market, how often you are recommended, what the models look up, whole buying conversations, your reputation and what the market says. If you want to know why AI does not recommend your brand, you need the AI Search Strategy.

Read more

Can I start with the AI Search Website Audit and do the AI Search Strategy later?

Yes. The AI Search Website Audit is part of the AI Search Strategy, so within 60 days of your audit's delivery you only pay the difference: 1,000 euro. We build on your audit and you get the Strategy within two weeks. After 60 days we first measure your site again, so the whole strategy rests on one measurement moment. Then you pay the difference plus a re-measurement: 1,195 euro. You keep one re-measurement, the Strategy's. From the Strategy to Strategy & Workshop you also only pay the difference: 1,000 euro.

Read more

What is a re-measurement?

At a moment you choose, after you have put the recommendations in place, we measure the same topics on the same instrument and go through the outcome with you: what really changed and what can still improve. We only report a change when the bands of the two measurements no longer overlap, so you know the change is real. One re-measurement is always included in your package. After that, a re-measurement costs 195 euro after a AI Search Website Audit and 495 euro after a Strategy.

What is included in a workshop?

Everything in the AI Search Strategy, plus a half-day workshop with your team, on site in the Netherlands or online. Together we turn it into owners, priorities and a plan for the next quarter. Strategy & Workshop costs 2,495 euro, and you can also add the workshop to an AI Search Strategy later, in consultation.

Read more

Working with us

How do I get access to Recommendable?

Request your package directly through the form. Then you fill in your brand profile through the intake form, we go through it together in a short call and we run the audit. If you want to get acquainted first, book a half-hour conversation through the contact page.

Read more

What do you need from us for an audit?

Your brand profile, which you fill in through the intake form, and a short call to go through it together. For a workshop, also your team for half a day.

What do I get after an audit, and how fast?

Within a week of the intake (AI Search Website Audit) or two weeks (AI Search Strategy) you get an advice report, a presentation and an action plan sorted by priority, with an owner for each action. For the AI Search Strategy, Peter goes through the key findings with you in a one-hour personal presentation. You can hand the action plan straight to your team or agency. We agree the date of the workshop together.

Does a person read the results?

Yes. One of the founders, Peter Minkjan or Erik Mus, reads every finding, writes the conclusion and builds the presentation. Nothing goes out unread, which is why an audit takes one to two weeks.

What does the workshop look like in practice?

The workshop takes half a day, on site in the Netherlands or online. Invite the people who will pick up the actions (brand, content, PR and tech) and whoever decides. We prefer to schedule it within two weeks of delivery, so everyone has read the report.

Read more

Data and security

Where does Recommendable store our data?

Your data is stored with Supabase (database) and Vercel (hosting), both in Stockholm, Sweden. Email is sent through Resend from Ireland. AI analyses go through the Vercel AI Gateway to Anthropic, OpenAI, Google and Perplexity, which may process data outside the EU.

Read more

Which AI providers see our data?

All AI calls go through the Vercel AI Gateway to Anthropic, OpenAI, Google and Perplexity, on paid business accounts. The gateway does not retain prompts and responses after the call. We only send what the analysis you asked for needs.

Read more

Who can see our data in Recommendable?

Your brand profile and your results stay with us and we share them only with you. Only the two founders have access to the production environment, with two-factor authentication. They also read your audit to write the advice.

Do you use our material in publications?

No, not without your permission, and we do not share client material with others. If you want to sign a non-disclosure agreement (NDA) before we start, that is possible.

How do I have my account and data deleted?

Email a request to info@recommendable.co. We delete your account within 7 days, except for data we must keep for legal or accounting reasons. The daily database backups still hold the data for at most 7 days after deletion.

How do you test security, and what if something goes wrong?

We test security ourselves, at least every quarter and before every major release. If there is a data breach, we notify you within 48 hours.

Read more

Agencies

Does Recommendable work for agencies and consultants?

Yes. Agencies and consultants use the technology itself: unlimited users and brands, organisations and roles per client, and the presentation deck under your own name. The price depends on how many brands you run and how often you measure, and we discuss it after a demo.

Read more

We already have an SEO, content or PR agency. Does Recommendable still fit?

Yes. The action plan is written so an existing team or agency can pick it up straight away, with an owner for each action. You can hand it straight to your agency.

Ready to start?

Request your package directly and we plan the intake together.

Request now

Fixed price · no subscription · results within two weeks