See how AI describes your brand to the customers you actually want.
We ask ChatGPT, Perplexity and Google AI Overviews the questions your customers are already asking, and report back what they say about you.
Your customers have started asking AI what to buy.
Say someone with $10,000 in savings asks ChatGPT which investing app to use. It doesn’t hand over a list. It names two or three, explains who each one suits, and leaves everyone else out without ever saying so.
You see none of that. No referral log, no report, no record the conversation happened at all.
So we go and have it. Thousands of times.
Here’s what comes back.
Case study: Robinhood
How often Robinhood is named, situation by situation. Pick one.
I’ve never invested before. Where do I even start?
What AI says backIn the engine’s own words“Many beginners start with Robinhood for its simplicity, then add another broker as their goals evolve.”
Perplexity
When people ask where to open a first account, Robinhood is the usual answer. There is no minimum to open one, trades are commission free, and fractional shares start at $1. It is sometimes described as the least intimidating place to buy and sell. Almost all of the praise is about how easy it is to start, and very little of it is about investing. Robinhood has no mutual funds and no bonds, and this is often listed as a reason to choose another broker from the beginning. There are no screeners and little educational material, and beginners who want to learn are usually directed elsewhere. The same app also puts options, crypto, margin and futures next to the buy button, which comes up repeatedly as a risk for a first-time investor.
I want to buy individual stocks and build my own portfolio, not just sit in index funds.
What AI says backIn the engine’s own words“Very simple, beginner-friendly interface for buying individual stocks and fractional shares, though it lacks the deep fundamental research tools of Fidelity or Schwab.”
Google AI Overviews
The three things most often praised here are the same as for a first account: it is easy to use, it is cheap, and it has fractional shares. The difference is that here they come with a limit attached. Robinhood is recommended “if simplicity matters most”, and the same answers then point to Fidelity, Schwab or Interactive Brokers for anything beyond that. What is missing is the material someone uses to pick a stock. There is no advanced screening, no deep charting, no fundamental data and no historical financial statements, and the screening that does exist requires a paid Gold subscription. Robinhood is also used as a point of comparison in descriptions of other brands. Webull’s research tools, in one of them, make it “much more powerful than minimalist apps like Robinhood.”
Which brokerages are offering the best bonuses for transferring a 401(k) right now?
What AI says backIn the engine’s own words“You must keep the matched funds in the Robinhood IRA for 5 years, and maintain your Gold subscription for at least 1 year, or face early removal penalties.”
Google AI Overviews
Almost four answers in ten bring up the match. It is 1% on a rollover with no cap, and 3% on contributions for Gold subscribers. It is described as the best offer available, because competitors cap theirs. The conditions attached to it are almost always listed alongside it: the matched money has to stay for five years, Gold has to be held for a year, and transferring out later costs $100. Robinhood is ranked first on the offer, and the same answers then compare it unfavourably to Fidelity, Schwab and Vanguard on everything else a retirement account has to do. Because there are no mutual funds there are also no target-date funds, so the mix someone already held has to be rebuilt in ETFs.
I want one login for everything. How do I bring my accounts together?
What AI says backIn the engine’s own words“Robinhood is optimized for bringing investments into Robinhood, while the $100 ACATS-out fee makes it less flexible if you later decide to move elsewhere.”
ChatGPT
The praise here is about moving accounts in. The transfer itself is simple, Robinhood reimburses up to $75 of the old broker’s fee, and one app holds taxable, joint, IRA and crypto accounts together. That last point barely comes up in any other question. The difficulty is with what cannot be transferred. Mutual funds and bonds have to be sold first, and any capital gains fall on the investor. Transferring out later costs $100. What Robinhood is best known for barely registers here: being easy for beginners comes up three times, against 95 times in the first-account question.
Each arrow is a dimension, and a brand is pulled towards the ones it is most described by. A brand in the middle is described by all four about equally.
The pages describing you aren’t the pages you wrote.
When AI names Robinhood, it reads Robinhood’s own site 11% of the time. Fidelity, Schwab and Vanguard all sit between 24% and 42%. So when an engine explains what Robinhood is, it is almost always using someone else’s description of it.
Three publishers decide what AI says about you.
NerdWallet, Forbes and StockBrokers.com produce 61% of everything negative AI says about Robinhood, concentrated in ten pages. Forums and social media account for 3%.
AI is reading your FAQs, not your product pages.
Two thirds of what AI cites from Robinhood is support content. For Fidelity it’s 7%, and the rest is product and editorial pages. Support pages exist to spell out what a product does and doesn’t do, so they’re full of limits, and the limits are what gets quoted. Robinhood’s own page listing what you can invest in is cited as proof that there isn’t much.
The AI models didn’t invent those words.
Almost every one of them already exists as text somewhere: your own pages, comparison articles, forum threads, reviews, Reddit conversations. The models read it and draw conclusions.
So we trace it back to the source. For each one, we show you which sources are doing the work, which turns a vague problem into a specific one. These pages. This wording. This gap.
What happens next depends on you. Some clients hand it to their content team, some brief their agency with it. Others use it to settle an internal argument about who their product is really for.
The questions come from real people.
We start with real customer language: forums, communities, reviews, the places people in your category already talk. That’s where the questions come from, in the words people actually use, including the complaints and warnings nobody phrases as a question.
Those questions go to ChatGPT, Perplexity and Google AI Overviews, repeatedly, over a week.
We want to get the questions right. It is the part the tools and agencies tend to neglect, and it decides whether any of the findings are worth acting on.
What the study tells you
- 01Where you appearWhich questions and which customer situations bring you into the answer at all.
- 02How you’re describedThe attributes, strengths and limitations the models repeatedly attach to your brand.
- 03Who you’re being sold toThe customers AI treats you as a fit for, and the point at which that changes.
- 04How competitors are framedNot just who else appears, but the distinctions the models draw between you.
- 05Which sources shape itThe pages the models lean on, so you know where the description is coming from.
- Talk to us
Three studies. Each one adds a layer.
Numbers only
“How visible are we?”
- Your visibility score, for each kind of customer
- Who AI names ahead of you
- How favourable the answers are
- Your top strengths and your top objections, each with a figure
Why AI says any of it, or anything about your competitors beyond where they rank.
Talk to usNumbers + qual analysis
“How does AI see us?”
- Our analysis of how AI describes you in each customer situation, and the reasons it gives
- What AI is actually pointing at when it says “easy to use”, or “not for serious money”
- The same analysis for every competitor
- A map of where each brand sits, and which part of your position is only yours
+ AI source analysis
“Where does it come from, and what do we change?”
- Which page produced each thing AI says about you
- Which publishers write your criticism, and which ones matter most
- How much of your description comes from your own site
- A ranked list of the pages worth changing, and what each should be doing instead
Final pricing depends on the number of competitors in scope.
Questions people ask us.
Is Rare Signal a tool?
There is no Rare Signal platform to log into. You give us the question, we do the collection and research, and you receive the finished study.
How does a project work?
Three stages, from question to answer.
The category, how you see your customers, and any research you already have. The last part is optional, but it shapes the questions.
We research how people in your category actually talk about the problem, and the questions come out of that.
Each question is asked repeatedly across a week, so every figure is a share of all the answers rather than a reading of any one of them. On the source study we also come back with what we think should change, and work out what is possible with you.
Can we see the questions before you run them?
We share the question set for sign-off before anything runs, and you keep it afterwards along with the method.
Which AI systems do you cover?
ChatGPT, Perplexity and Google AI Overviews, asked logged out, the way a customer with no account history would see them.
Which markets can you cover?
Any market where your customers ask in English or French. The engines are asked from that country, so a UK study is asked from the UK. We do not work in other languages, because the analysis depends on reading the answers rather than counting them, and that only works in a language we read properly.
How long does it take?
Two to six weeks, depending on the study and the number of competitors in scope.
Start here
Researcher
Media buyer
New business
Find out what AI is tellingyour customers.
Tell us the category and who you compete with. We’ll tell you how we’d approach it.