Part 1: technology – without AI queries
We fetch your website the way a visitor or crawler does and check whether it offers good conditions for AI assistants. This covers:
- Reachability: may and can the AI providers’ search crawlers read your pages – including past your host and bot protection?
- Unambiguous facts: are business, services and contact details machine-readable and written the same way everywhere?
- Content: are there clear pages for your main services and your clients’ typical questions?
- Basics: sitemap, load time, server stability.
The result is a technical score from 0 to 100 with a traffic light per area. We go easy on your server: if we notice a block, we stop.
Part 2: AI visibility – 20 questions to 4 assistants
- Questions per client: we compile the questions with you – the way your clients ask, matching your industry, services and region. You approve the questions.
- Same conditions: we query ChatGPT, Gemini, Perplexity and Claude through their official interfaces with web search. Language and location are documented, every answer is stored with its date.
- Evaluation: is your business named, how is it described, which sources do the assistants cite – and how do your 3 competitors fare? We also compare the details in the answers with your real data. We flag deviations as “please check”, not as “wrong”.
- Verifiable: every number in the report is based on stored answers. How the AI score is calculated is stated in the report.
Before and after
After implementation we put the same questions to the same assistants and compare with the last result. We do not change the questions after that, otherwise the comparison would be skewed.
Besides mentions, what matters to you is what comes of them: a mention is not yet an enquiry. Where possible, we therefore also evaluate visits and enquiries.
Why not 100?
A technical score below 100 is not automatically a flaw. Some checks do not fit every business: an online service without a shop needs no opening hours, and in countries without postal codes we do not ask for one. We explain open points in the report under “Why not 100?”.
Our own website pickedfirst.ai reaches 100 out of 100 – without made-up details. For Unser Dubai, 5 points were missing at the end because our measurement looked for different service terms than the website’s.
No made-up details
We do not enter a postal code “00000”, round-the-clock opening hours or a local-shop entry just to gain points. Made-up details would mislead AI assistants and harm you. And we do not invent reviews, client quotes or numbers – neither for you nor for ourselves.
Limits of the check
- AI answers fluctuate. The same question can get a different answer minutes later. The check is a sample at one point in time, not a permanent ranking.
- Answers in the apps (ChatGPT, Claude.ai, Gemini app, Perplexity) can differ from the interfaces, e.g. through personalisation, location or other model versions.
- Technical ratings are professional judgements, not an official formula of the AI providers.
- Nobody can promise that an AI will name your business. The measures improve the conditions.
Which data goes to the AI providers
Only public company data in the form of questions: company name, place, industry, services and the names of competitors. No personal data, no client data.