Documentation
Reading your results
Picked shows a lot of numbers. Three of them tell you most of what you need: your visibility score, who is named instead, and which pages are cited. This page explains those and the rest.
What each number means
| Number | What it counts |
|---|---|
| Visibility score | Out of all the answers in the date range, the share that name you. The number to move and to put in a report |
| Rank | Your place among every brand the answers name, by visibility score. "#4 of 31" means three brands are named more often |
| Avg. Position | Your average place in the answers that name you. 1 means you're named first |
| Sentiment | How the answers that name a brand present it, 0 to 100: 0 warns against it, 50 is a neutral mention, 100 recommends it first |
| Cited | An answer links a page on your website as a source |
| Demand | Google searches a month in the US for the closest search to the prompt |
Changes, the small green or red figures, compare the date range with the same number of days just before it. A dash means there is nothing to compare with yet.
Named or cited
These are different, and both matter.
- Named means your brand appears in the answer's text: "Try Acme, HubSpot or Pipedrive."
- Cited means the answer links one of your pages as a source.
An answer can name you without citing you (the model knows you, from training or from other people's pages), or cite your page while recommending someone else. Being cited is how a page you wrote gets into the answer.
How Picked decides you were named
An answer counts as naming you if it uses your brand name, your domain or any of the other names in Settings, General. If an assistant calls you by a product name, add it there, or those answers count as not naming you.
Competitors work the same way: each tracked brand has its own other names, counted as the same brand in the rank.
A real first result
We run a launch platform called LaunchOnIt. On September 27 and 28, 2026 we asked ChatGPT, Gemini, Perplexity and Google AI Overviews five questions a founder would ask about where to launch. That is 40 answers.
| What we counted | Result |
|---|---|
| Answers | 40 |
| Brands named across them | 76 |
| Answers naming LaunchOnIt | 0 |
| Answers naming Product Hunt | 33 (83%) |
That is a normal first result, and a useful one. Three things to take from it.
Zero is a starting line
New brands are usually named in few answers. What matters is whether the number moves, not where it starts.
Look at who is named
Product Hunt, BetaList and Indie Hackers came up again and again. Those are the names the models trust for that question. If you are a newer competitor, this is the list you are up against. Brands shows the ones you haven't added yet under "Named by the AI models, not tracked yet".
Look at the sources
Open a prompt and read which pages were cited. Those are the pages the model read before it answered. To be named, you usually need to be on those pages, or to have a better page for the same question.
Why answers change
Ask the same question tomorrow and you can get a different answer: in our ChatGPT answers, fewer than half the pages cited one day came back the next. Answers also differ between models and between countries. The full numbers are in the ChatGPT citations study.
So don't judge a prompt by one day. Judge it by the trend over a few weeks, and by your share across many prompts. That is why Picked asks every day, and why one screenshot proves little.
Why the numbers don't match
A few things make two numbers disagree, and none of them is a bug:
- Models see different pages. ChatGPT, Perplexity and Google's AI answers each search their own way, so a brand can lead on one and be missing on another. Compare them on a prompt's page, model by model.
- Claude is asked less often than the others. Over a short range it has fewer answers than the others, so one answer moves its share more.
- Small ranges swing. Five prompts on two models over a week is a few dozen answers. A single answer can move the score by several points. Widen the date range before you read a trend.
- A new prompt starts from its first answer. Adding prompts mid-month changes the mix the score is averaged over.
- Demand is a Google figure. It says how many people search something close on Google in the US, not how often people ask an AI. Use it to rank prompts against each other, not as a count of AI questions.
A prompt you lose: what to do
- Open the prompt and read the answer. Who is named, and why? The answer often says: cheapest, best for agencies, most integrations.
- Read the sources. Which pages did the model cite, and what do they say about the brands named?
- Get onto the pages that exist. If a list or review site named your competitors and not you, getting listed there is often the quickest gain. Get listed in Content Gaps collects these pages across all your prompts.
- Write the page that doesn't exist. If the answer should have been a page on your site, start an article from Content Gaps. See write the article.
- Watch the prompt. Picked keeps asking it every day. Give a change a few weeks before judging it.
What to report
For a client or a boss, three numbers carry the story: the visibility score and how it moved, the rank against the brands they care about, and one prompt won or lost with the answer that shows it. On Scale, client reports put these in a PDF under your name.
Next
AI crawlers shows the other half: which of your pages the AI opens.
Updated
