Who does AI recommend today?
Neutral buyer questions in the local language on ChatGPT, Google AI Mode, Perplexity and Copilot.
- Visibility and share of voice
- Who is named first
- The sources AI cites
Your market, as AI reads it.
Buyers, analysts and advisers ask AI first. We measure what ChatGPT, Gemini, Claude, Perplexity and the agents that browse for them say about every company in a market, how accurate it is, and what they can read on each company’s website.
57%of web page requests now come from bots, not people.Cloudflare, June 2026 ↗
What is an AI Perception Study?
An AI Perception Study is commissioned research by Reaudit that measures how AI systems see one industry in the markets you choose. For every company we record what ChatGPT, Google AI Mode, Perplexity and Copilot recommend, what GPT, Claude, Gemini, DeepSeek, Mistral and Grok believe from memory, how accurate it is, how AI describes the market itself, and whether AI agents can read each company’s website. You receive an interactive report, a PDF and the full dataset.
Behind the page · reaudit.io
People see the design. Agents read the source.
This is our own homepage. Move over it: the lens shows the HTML that GPTBot, ClaudeBot and PerplexityBot receive, before any JavaScript runs. Every line is real, checked against the live site.
01 / 07The headlineread by agents"Make your brand part of the answer." is plain text in the h1, so every crawler gets it.
98.5% of the words a person sees on reaudit.io are in the HTML agents receive. Most companies in a study are far lower: prices, ratings and buttons that only exist after scripts run. That is what we measure, company by company.
2,392words a person sees on the homepage
2,356of them in the HTML GPTBot receives
What a study measures
Four ways AI sees a market
One industry, the markets you choose, 30 to 200 companies. The same questions everywhere, every answer kept, so every number can be checked.
What do the models believe?
The same questions to GPT, Claude, Gemini, DeepSeek, Mistral and Grok with web search off.
- Each company’s profile
- Accuracy against its own facts
- The words and tone used
How does AI describe the market?
Size, structure, leaders, trends and risks, as each model explains them to a buyer.
- Where models agree and differ
- Local language vs English
- Known, but no longer recommended
Can agents read each website?
Every company’s site, read the way AI crawlers and browsing agents read it.
- AI bots blocked, and by what
- Content that needs JavaScript
- Prices, products and contact readable
From a sample report
One company. Six models. Six stories.
Each model answers the same profile questions with web search off. Every answer is checked against the company’s own facts and quoted word for word, with model version and date.
Halcyon Mutual, as each AI believes it
Illustrative data| Fact |  GPT |  Claude |  Gemini |  DeepSeek |  Mistral |  Grok |
|---|---|---|---|---|---|---|
| What it does | Correct | Correct | Correct | Life only | Correct | Correct |
| Founded | 1994 | 1994 | 2001 | Doesn’t know | 1994 | 1989 |
| Headquarters | Athens | Athens | Athens | Thessaloniki | Athens | Doesn’t know |
| Chief executive | Former CEO | Correct | Former CEO | Doesn’t know | Former CEO | Correct |
| 2025 merger | Knows | Knows | Doesn’t know | Doesn’t know | Doesn’t know | Knows |
| Described as | Reliable | Reliable | Expensive | Small | Reliable | Slow claims |
CorrectOutdatedWrongDoesn’t know
“Describe the motor insurance market”
Illustrative dataGPT web search offA concentrated market led by three large insurers, with online brokers growing fast among younger drivers.
Gemini web search offA concentrated market where bancassurance sells most new policies, and prices rose sharply after 2023.
Claude web search offThree insurers hold most of the market; comparison sites and brokers are where buyers start.
Blue: what the models agree on. Pink: a claim only one model makes.
Findings you can cite
Does being readable by agents pay off?
Every layer covers the same companies, so a study answers questions no single tool can. Do agent-ready websites get recommended more? Do companies AI describes wrongly lose visibility? Who leads in the models’ memory but loses in live search?
Findings are computed from the data. A model never adds a number.
Agent readiness vs AI visibility
Illustrative dataWho commissions a study
For firms that sell what they know about markets
Expert networks
A sector briefing before client calls: who AI favours, and what it gets wrong.
Consultancies
Proprietary data for strategy, due diligence and digital work, licensed for client deliverables.
Business information providers
A new dataset next to credit and company data: how AI sees every company in a sector.
How it works
Ten working days, scope to report
- Scope
Industry, markets and companies, or we propose the list.
Day 1 - Ask and read
Every engine and model in each market’s language; every website read as agents read it.
Days 2–6 - Check
Answers graded against company facts; quotes verified word for word.
Days 7–8 - Deliver
Interactive report, PDF and dataset, and a call to walk your team through it.
Days 9–10
What you receive
A report to read, data to reuse
Every study is scoped to your industry, your markets and the companies that matter to you.
- Interactive report
- A private page with every chart, company profile and source view.
- PDF report
- Reaudit, co-branded, or under your own name.
- Dataset
- Excel and CSV: every answer, grade and website check.
- Method appendix
- Questions, engines, model versions, dates and locations.
- Walkthrough call
- We take your team through the findings.
- Quarterly refresh
- The same study again, to show change over time.
Tell us the industry and the markets. We’ll send a proposal.
A company list, the markets, the delivery date and the cost, within one working day.
Questions
Frequently asked
What is an AI Perception Study?
Commissioned research that measures how AI systems see one industry in the markets you choose: who AI search recommends, what AI models believe about each company and how accurate it is, how AI describes the market, and whether AI agents can read each company’s website.
Which AI engines and models do you cover?
AI search: ChatGPT, Google AI Mode, Perplexity and Copilot. AI from memory, with web search switched off: GPT, Claude, Gemini, DeepSeek, Mistral and Grok. Every answer is stored with the model version, date and location.
How long does a study take?
About ten working days from an agreed scope to the delivered report, PDF and dataset.
Can we publish or resell the results?
Yes, with a white-label licence. The report and data can carry your name, be co-branded, or stay as Reaudit.
How do you check that AI answers are accurate?
Each answer is compared with the company’s own published facts and graded correct, outdated, wrong or unknown. Every quote in the report is checked word for word against the stored answer.
How is this different from Reaudit’s public industry studies?
The public studies show which brands ChatGPT and Google AI Mode name in an industry. A commissioned study covers the industry and companies you choose, adds what AI models believe from memory, how accurate it is, how AI describes the market, and whether agents can read each website, and comes with the full dataset.
How much does a study cost?
It depends on the number of markets and companies. Tell us the industry and the markets and we send a proposal with the cost within one working day.
Commission a study
Tell us the industry and the markets
We reply within one working day with a proposed company list, a delivery date and the cost.