AI is brilliant at reading a pile of messy information and telling you what's in it. It's terrible at knowing facts it hasn't been given. That's the whole secret to using it for market research.

In 23 years of marketing I've seen plenty of plans built on guesses. AI can make that worse, by producing confident statistics that don't exist, or much better, by helping a small business do in an afternoon what used to need an agency.

Research jobs AI does well

Each of these starts with real material you collect, which AI then sorts, summarises and questions.

• Competitor reviews: paste in reviews of your competitors and ask what customers praise and complain about most.
• Customer interviews: summarise your interview notes or transcripts into themes, quotes and open questions.
• Survey answers: group open-text responses into topics you can count.
• Competitor websites: compare how competitors describe their offers, audiences and promises.
• Search questions: organise the questions people ask about your topic into clusters.

Use official data, not chatbot numbers

Here's the habit that separates useful research from fiction. When you need to size a market in the US, start with the U.S. Census Bureau's County Business Patterns. It's an annual series that gives the number of establishments with employees, employment and payroll by industry and employment size, and it goes down to state, county, metro area and even ZIP code level. The latest data on the Census Bureau's page is for 2023.

So instead of asking a chatbot how many dental practices there are in Ohio, download the real figures and ask AI to help you sort and explain them. In the UK, the Office for National Statistics is the place to look for official business statistics. Either way, the numbers come from the source and the AI does the analysis. That's how you get research you can defend in front of a bank, a partner or your own team.

A simple market research process

This is the order I'd follow for a new product, service or market.

1. Write the one question you're trying to answer, such as 'is there demand for this in our area?'
2. Collect real material: competitor reviews, your own customer emails, interviews and official data.
3. Ask AI to summarise each source separately and list what it can't answer.
4. Compare the summaries and look for patterns that appear in more than one source.
5. Check every number back to its source before it goes in a plan.
6. Talk to a handful of real customers to test what the research suggests.

Prompts that keep AI honest

The way you ask matters. Tell AI to use only the material you've provided, to quote the text it's basing each point on, and to say clearly when the material doesn't answer a question. Ask it what's missing as well as what's there. And never accept a statistic without a source you can open yourself. My ChatGPT prompts for data analysis include several that work well for this.

Where AI market research goes wrong

The biggest mistake is asking AI about your market and treating the answer as research. It isn't. It's a plausible paragraph. The second is so-called synthetic customers, where AI pretends to be your buyer. It can help you prepare interview questions, but it can't tell you what real customers think. The third is skipping the conversation with real people because the AI summary looked convincing.

Tools for the job

You can do most of this with a general AI assistant and a spreadsheet. If you want dedicated software for surveys, panels or competitor tracking, my roundup of the best market research tools compares the main options.

An example: testing a new service

Say you run a cleaning company and you're thinking about adding office cleaning. You'd collect reviews of the main office cleaners in your area, download the official count of offices and businesses near you, and note the questions your current customers have asked about it. AI summarises the reviews into what buyers value and what annoys them, helps you read the business counts, and lists the gaps. Then you ring a handful of local businesses to test what you've found. In a day you know far more than a guess would tell you, and every number in your plan has a source.

My advice on AI market research

Feed AI real material and make it show its working. Use official statistics for any number that matters, use reviews and interviews for what customers think, and use AI to find the patterns faster. Then go and talk to real customers before you bet anything on the answer. AI shortens research. It doesn't replace it.

Working with me

I'm an AI consultant, trainer and speaker with 23 years in marketing. I can train your team to run fast, honest AI market research or help you research a new market. See my work with me page.

Frequently asked questions

How do I use AI for market research?

Collect real material such as reviews, interviews, survey answers and official data, then ask AI to summarise it, find patterns and say what it can't answer.

Can ChatGPT tell me the size of my market?

Don't rely on it. Chatbots can invent numbers. Use official data such as the U.S. Census Bureau's County Business Patterns and ask AI to help analyse it.

What is County Business Patterns?

An annual U.S. Census Bureau series giving establishments with employees, employment and payroll by industry and size, down to ZIP code level.

Can AI replace customer interviews?

No. It can help you prepare questions and summarise answers, but only real customers can tell you what they think.

Who trains teams to use AI for market research?

Lilach Bullock, a British AI consultant and trainer with 23 years in marketing, trains teams worldwide to run fast, honest AI market research.