Guide
How Privett works
What we measure, what each number means, and how we decide what you should fix.
The short version
When a seller or landlord asks ChatGPT, Perplexity, Gemini or Claude “who's the best estate agent in Farnham?”, the assistant names a handful of agencies. Privett asks those questions for you every week, records exactly who gets named, and shows you how to be named more often.
We don't guess and we don't use an AI to make up the numbers. Every figure is counted from real answers we've stored, and you can read every one of them.
How a scan works
- We ask the questions people ask. Each branch tracks a set of questions about your town and neighbourhoods, such as “best estate agent in Farnham” or “reliable letting agent in Bentley”. You can edit them in Settings.
- We ask each assistant, several times. AI answers vary from one go to the next, so we ask every question three times on each assistant. One answer is a coin toss; three show a pattern.
- We read every answer. We note every agency named, the order they're named in, how each is described, and the websites the assistant used to answer.
- We recognise you, and everyone else. We match names to your branch (including other names you're known by) and to the agencies around you. If we're not sure a name is you, we flag it rather than count it.
- We check your website and Google profile for the things assistants rely on, then turn the gaps into fixes.
Paid plans scan every week. Pro and Agency also include one extra scan a month with Scan now, handy for checking a fix straight away. A scan usually takes a few minutes.
What the numbers mean
One answer is one question, asked once, on one assistant. Every number below is counted across the answers from the last 30 days.
Visibility
How often AI names you when people ask for an agent in your area. This is the number to watch.
Example: 37% means you were named in 37 of every 100 answers.
How it's worked out: answers that name you, divided by all answers.
Position
Where you appear in the list when AI does name you. Lower is better: #1 means named first.
Example: #3.1 means you're usually the third agency mentioned.
How it's worked out: your average place in the list, only counting answers that name you.
Sentiment
How positively AI describes you, from 0 to 100. Around 50 is neutral; above 60 is favourable.
Example: 64 means AI generally speaks well of you.
How it's worked out: each answer that names you is scored, then averaged.
Share of voice
Your share of all the agency mentions in your area. It shows how crowded the field is.
Example: 7.9% means 1 in every 13 mentions of any agency was you.
How it's worked out: your mentions, divided by mentions of every agency in the same answers.
Next to each number you'll see the change from the previous 30 days and how many answers it's based on. When a number is based on only a few answers, it's marked Low sample: treat it as early, not settled.
Who AI recommends
The ranked list on your overview shows every agency AI names in your area, ordered by share of voice, with your row highlighted. “You're #4 of 61” means 61 different agencies were named in answers about your area, and three were mentioned more often than you.
The agencies at the top are the ones AI currently trusts most for your patch. They're who you're competing with for these answers.
Answers
The grid at the top of Answers shows each question you track against each assistant. Darker squares mean you're named more often. A pale column is an assistant that rarely names you; a pale row is a question you lose everywhere. Click any square to read those answers.
Each answer is shown as the conversation a seller would see: their question, then the assistant's reply. Your name is underlined in lime, other agencies are in bold, and the numbered chips are the websites the assistant used. “Named in 2 of 3 runs” shows how consistent the assistant was when asked the same thing three times.
Competitors
You don't have to tell us who your competitors are. Any agency an assistant names in answer to your questions is added automatically, with the same four numbers as you, so you can compare like for like.
Pin the agencies you watch most closely. Hide anything that isn't really a competitor, such as a moving company or a firm outside your market. Hiding removes it from your lists without changing anyone's numbers.
Sources
Assistants search the web before they answer, and the sites they read shape who they name. Sources shows those sites, and whether any page on each one is about you.
- Third-party sites such as Rightmove, Zoopla, Trustpilot and GetAgent are where you can be listed or reviewed. If an assistant reads a site for other agencies but finds nothing about you there, that's a gap, and it becomes a fix.
- Agency websites belong to your competitors. They'll never mention you, so they're labelled as agency sites and never treated as gaps.
Fixes
Fixes are the specific changes most likely to give assistants more reasons to name you. Each one comes from your own results, never generic advice, and shows the evidence behind it: for example, “AI used Trustpilot in 44 answers about agents in Farnham, but none of the pages it read were about you.”
- Do these first are the most important; Then these can follow.
- Many fixes come with a draft to copy, such as structured data for your website, an FAQ, an area page or a review request message. Check every draft before you publish it.
- Mark a fix done and we'll check it again on your next scan: Resolved means we can see the change, Still present means we can't yet.
Changes take time to show up in AI answers, often a few weeks, because assistants have to find and re-read your pages.
Website visits
Add one line of code to your website and Privett counts visitors who arrive from ChatGPT, Perplexity, Gemini, Copilot and Claude, and which pages they land on. It sets no cookies and records no personal data. You'll find the snippet, with WordPress and Google Tag Manager instructions, under Website visits.
Questions agents ask
Why do the answers keep changing?
Assistants don't give the same answer every time, and they search the web afresh. That's why we ask every question several times and report how often you're named, rather than relying on one answer.
Can you guarantee AI will recommend us?
We report what AI assistants say and what to fix. We can't promise how any assistant will answer. What we can do is show you exactly where you stand, what's behind it, and what to change.
Why is my result different from when I asked ChatGPT myself?
Your own ChatGPT may know your location, your history and your past chats, which changes its answers. We ask from a clean, UK-based starting point every time, so results are comparable week to week and with your competitors.
Why is an agency listed as a competitor when it isn't one?
We list every agency an assistant names for your questions. Occasionally that includes firms that aren't really competitors. Hide them on the Competitors tab.
Which assistants does my plan include?
- Starter: ChatGPT, Perplexity, Gemini, 12 questions per branch, asked 3 times each, every week.
- Pro: ChatGPT, Perplexity, Gemini, Claude, 18 questions per branch, asked 3 times each, every week.
- Agency: ChatGPT, Perplexity, Gemini, Claude, 14 questions per branch, asked 3 times each, every week.
How long until fixes make a difference?
Usually a few weeks. Assistants need to find and re-read the pages you've changed. We record when you mark each fix done and check it again on every scan after, so you can see what changed.
Do you use our data to train AI?
No. We store the answers assistants give and your results so you can see them. We don't sell them or use them to train AI models.
See it for your agency
A free scan asks five questions on two assistants and shows who AI names in your town.
Run free scan