How to Find Business Ideas with AI (Step by Step)

To find business ideas with AI, don't ask ChatGPT or Claude to invent them. Connect the assistant to real search data, ask it for niches that are growing, make it check monthly volume and two-year growth, cut anything seasonal, declining or tied to one brand, then look at the bigger trend behind what is left. Five steps, about 20 minutes, and every idea you end with has a number and a month attached.

Let's get into it.

Why asking ChatGPT for business ideas is not enough

Ask a chatbot "give me 10 business ideas" and you get a tidy list. The problem is where it comes from. The model is drawing on its training data, so it cannot tell you whether people searched for that thing more this year than last year. It also can't tell a steady market from a one-week TikTok spike.

Most pages ranking for this question are "AI business idea generators": you type your interests and skills, and a model returns ideas. Useful for brainstorming, but none of them checks demand.

The fix is to give the assistant a data source. The Model Context Protocol (MCP) is an open standard that lets AI apps like Claude and ChatGPT connect to outside tools and data. The MCP documentation calls it a USB-C port for AI applications. Anthropic introduced it in 2024, and both assistants now support remote MCP servers.

Rising Trends runs one. Once you connect the Rising Trends MCP for Claude and ChatGPT, the assistant can pull absolute US monthly search volume, five years of history and growth over 3 months to 5 years for 65,000+ tracked trends, instead of guessing. That is the whole trick: the model does the thinking, our database supplies the numbers. Other trend data servers exist too, and our ranking of the best trends MCPs and Claude skills compares them, including what each one does that we don't.

Set up the Rising Trends MCP (2 minutes)

You need a Rising Trends Pro account. As of November 5, 2026, Pro on the pricing page is a one-time $97 payment for lifetime access, and it includes the REST API and the Claude and ChatGPT connector.

In Claude:

1. Open Settings, then Connectors, then Add custom connector.
2. Paste https://mcp.risingtrends.co/mcp.
3. Click Connect, sign in to Rising Trends and press Allow.

Claude's help center says custom connectors using remote MCP work on Free, Pro, Max, Ultra, Team and Enterprise plans, with Free users limited to one custom connector.

In ChatGPT: add a custom MCP server, paste the same URL, choose OAuth, sign in and press Allow. Then mention @Rising Trends in your message.

In Cursor or Claude Code: create an API key in your account and send it as Authorization: Bearer rt_live_... with the same URL.

Here is the setup page, with the steps for each app:

Once the connector shows as connected, ask your first question. Everything below is a real session: real prompts, the real tool calls with their arguments, and answers built from what our database returned on November 5, 2026. The data runs through September 2026, and every volume is US monthly Google searches.

This card sums up the method:

Each step below follows the card in order.

Step 1: Ask for growing niches, with filters

Start wide but give the assistant hard filters. "Growing" means nothing without a window, and tiny keywords are noise. A floor of 2,000 monthly searches keeps the list to things real people look for.

Example conversation

Claude calls the find_trends tool twice:

find_trends { "type": "services", "timeframe": "1y", "min_volume": 2000, "sort": "growth", "limit": 40 }
find_trends { "category": "Home", "timeframe": "2y", "min_volume": 3000, "sort": "growth", "limit": 30, "page": 2 }

That is a list, not a decision. Several of those numbers will not survive the next step.

Step 2: Check monthly volume and 2-year growth

A one-year growth figure can come from a single month. So ask for the full history and the 2-year number, which is harder to fake. The get_trends tool returns up to 10 trends in one call, each with five years of monthly volume, growth over 3m, 6m, 1y, 2y and 5y, a seasonality label, momentum and whether it is a brand.

Example conversation

get_trends { "slugs": ["ai-answering-service", "technical-seo-agency", "hut-builders", "longevity-clinics", "gpu-repair", "interior-decorators", "cowork", "ai-call-centers", "privacy-email", "split-pay-rent"] }
get_trends { "slugs": ["terracotta-water-filter", "coloring-blanket", "bed-japan"] }

Read that list again. Only three climbed month after month. The chart for "ai answering service" shows what a real climb looks like:

Even if you ignore the August peak, the floor rose from 720 to around 2,000 over 18 months. That slow floor is what you want to see before you build anything. It is the same rule we use in our guide to trend forecasting: the baseline matters more than the peak.

Step 3: Cut seasonal, declining and single-brand terms

Now the cuts. Ask the assistant to apply them using the fields it already has, so you don't argue with a chart by eye.

• Seasonal: the seasonality field is medium or high, or the curve peaks in the same months every year.
• Declining: 3-month growth is negative and momentum is "slowing" or "peaked".
• Single brand: is_brand is true. You can't start a business selling someone else's product name.
• Spikes: one or two months carry the whole growth number.
• Not a solo business: regulated or capital-heavy, whatever the curve says.

Example conversation

get_trends { "slugs": ["catio", "dog-water-fountain", "cat-feeder", "dog-friendly-ice-cream", "litter-robot-5"] }

The chart below is why the spike rule exists. "Interior decorators" looks like the best idea on the list if you only read the growth column:

Three flat stretches near 1,300 searches, three walls. Whatever caused those walls, it is not a market you can plan around. For a deeper version of this check, see our walkthrough of trend analysis.

Step 4: Look at the mega trend behind the niche

Two survivors is too few, and a single keyword is fragile. The next move is to zoom out. Rising Trends groups related trends into mega trends: 70+ shifts in behaviour, each with combined search volume, sub-trends, and the websites and apps riding them. A niche that sits inside a growing mega trend has room around it if your first angle doesn't work.

Example conversation

list_mega_trends { "timeframe": "2y", "sort": "growth", "limit": 40 }
get_mega_trend { "slug": "ai-search-optimization" }
get_mega_trend { "slug": "trading-card-asset-management" }

This is where new ideas show up that a keyword list would never surface. "AI visibility audit" has only 480 searches, but it sits on top of a 246,000-search behaviour. Google rolled AI Overviews out to everyone in the US in May 2024, and the curve has climbed since. Small businesses now want to know if AI answers mention them. Our AI trends report tracks the wider category.

Step 5: Decide on 5 ideas backed by numbers

Last prompt. Ask for a short brief per idea, with the number, the month and the risk. Make the assistant say what would kill each one.

Example conversation

Here are the five side by side:

The chart shows that volume and opportunity are different things. The smallest bar, "ai visibility audit", is the fastest grower and has the biggest mega trend behind it. The coloring blanket bar is the biggest, but the history is the shortest. Pick by your skills and your risk appetite, not by bar length.

If idea 1 sounds familiar, it also leads our ranking of 25 AI business ideas by real demand. For more ideas outside AI, see the best businesses to start, ranked by rising demand. For idea 5, our look at which print-on-demand niches are rising covers suppliers and margins.

Copy-paste prompts

Paste these into Claude or ChatGPT once the Rising Trends connector is on. Swap the category for yours.

Using Rising Trends, find 30 trends in [category] with at least 2,000 US monthly
searches, sorted by 2-year growth. Pull the full history for the top 10. Remove
anything seasonal, declining over 3 months, tied to one brand, or driven by one or
two spike months. For what is left, find the mega trend each belongs to. End with
5 business ideas: the offer, monthly searches with the month, 2-year growth, the
mega trend, and the biggest risk.

Using Rising Trends, list the 20 mega trends that grew most over 2 years. Pick
3 where a solo founder could sell a service or a small product. For each, show the
5 smallest trends inside it that grew over the last year, and the websites already
serving it.

Using Rising Trends, look up [your idea]. Give me monthly searches, growth over
3 months, 1 year and 2 years, seasonality, momentum, whether it is a brand, and the
mega trend it belongs to. Tell me plainly if the curve is a spike.

What the AI still can't do for you

Search data shows demand. It does not show whether you can win. Three checks stay on your side of the screen:

• Competition. Search the term yourself and look at who ranks. Ask the assistant to run find_websites in the category to see which sites are growing.
• Willingness to pay. Searches for "ai visibility audit" mean curiosity. Five calls to local businesses tell you if anyone pays.
• Your fit. An idea with great numbers is a bad idea if you hate the work.

The US Small Business Administration's guide to market research and competitive analysis lists free government data sources for checking demand, costs and competition before you commit.

Also, data has a lag. Our numbers are monthly and the latest month is September 2026. That is a good fit for picking a business, where you care about direction over months. It is not a tool for chasing this week's viral moment.

How to repeat this every month

The method only works if you rerun it. Trends that pass the cuts today can stall in three months, and new ones appear. Two ways to keep it going:

• Rerun Prompt 1 at the start of each month for your category and compare it with last month's answer.
• Automate it. The same data is in the Rising Trends API, so a short script can pull your filters every month and flag new names.

You can also track up to 100 trends with Pro and get monthly search-volume alerts, so your shortlist tells you when it moves.

FAQ

Can ChatGPT come up with business ideas?

Yes, but on its own it works from training data, so it cannot tell you if demand is growing now. Connect it to a data source through MCP, like the Rising Trends connector, and ask it to check volume and growth before it suggests anything.

What is the best AI tool to find business ideas?

A general assistant (Claude or ChatGPT) plus a data connector beats a one-click "idea generator". Generators return ideas from the model's memory. With a connector, the same assistant can filter real search data, apply your cuts and explain its reasoning.

How do I validate a business idea with AI?

Ask for the monthly history and 2-year growth, check seasonality and momentum, and look at the mega trend around it. Then validate outside the chat: talk to buyers and check who already ranks. Our guide on how to use Google Trends to find a profitable niche covers the free version of these checks.

Do I need to code to use the Rising Trends MCP?

No. In Claude or ChatGPT you paste one URL and sign in. Only coding tools like Cursor need an API key in a header.

Want to spot the next shift before everyone else? Read our guide on how to identify market trends, or browse what is breaking out right now in our trending topics list.