Can You Really Do M&A Research With ChatGPT? We Tested It.

This is not a post about why ChatGPT is bad or why AI is overhyped. ChatGPT is a genuinely useful tool for M&A practitioners for certain tasks in certain contexts. This is an honest account of where it works, where it runs out, and what the difference means for teams using it as a research layer for live mandates.

We tested it. Here's what we found.

What you can actually do with ChatGPT

Market mapping at a broad level works reasonably well. If you need a starting list of companies operating in a given sector, ChatGPT can produce that list faster than a manual search, with the important caveat that it's drawing on its training data, not a live database.

Draft outreach is another genuine use case. For polishing messages, restructuring arguments, or adjusting tone, it performs well. It's less effective when generating outreach from thin context; the output tends toward generic when the input is generic.

Document summarisation and synthesis are legitimately useful. If you have an information memorandum or a buyer profile and need to extract key points quickly, ChatGPT handles this well. The output quality depends on the quality of the input, but for synthesis tasks it's a real time-saver.

Where it breaks down for M&A

Proprietary deal data is the first hard wall. ChatGPT doesn't have access to recent transaction data, fund portfolio updates, or private company financials. The buyer profiles it produces are built from public information that may be months or years out of date. A fund's stated thesis from a 2022 press release may not reflect where they're actually deploying capital today.

Structured financial screening isn't natively supported. Running a screen against EBITDA range, geography, and ownership type requires integration with data sources that ChatGPT doesn't connect to without significant custom tooling.

The validation layer is absent. When ChatGPT produces a list of potential acquirers, there's no mechanism to verify that the contacts are current, that the companies are actively acquiring, or that the financial profiles match what the screen specified. It generates plausible outputs, not verified ones. For low-stakes research tasks, that's acceptable. For building a buyer universe you're going to put a mandate on, the stakes are different.

What context scale means in practice

One of the less-discussed limitations is context volume. A general AI model handles a useful amount of context per session , but M&A mandate research involves cross-referencing dozens of data sources, validating outputs against multiple databases, and iterating across hundreds of counterparty profiles.

Purpose-built research engines process significantly more context per mandate than a general AI session can manage by cross-referencing sources, validating against deal history, and building personalisation at scale across large counterparty lists. That depth shows up in buyer list quality, not just generation speed.

The honest summary

ChatGPT is a useful research assistant for M&A , it is not a research platform. The gap is in three specific areas: proprietary data access, M&A-specific validation logic, and personalisation at scale.

For early-stage exploration, document synthesis, or outreach polishing, it's a reasonable tool that most practitioners should be using. For building a buyer universe you're going to deploy outreach on, run a mandate through, and present to a client, the bar is different and a general AI tool doesn't clear it.

The question isn't whether to use AI. It's which AI, for which tasks, with what data access behind it.

What purpose-built means

Purpose-built for M&A means the research logic is trained on deal-specific data , what makes a buyer "right" for a mandate , rather than adapted from general text. It means integrated data sources for financial verification. It means the validation layer exists before output reaches the user.

It also means the outreach output is designed for M&A buyers , speaking to acquisition history, portfolio logic, and deal fit, rather than generic B2B sales language adapted for finance.

The difference shows up most clearly when you compare outputs side by side: the buyer list from a general AI prompt versus the buyer list from a system that read the mandate, cross-referenced live data, and ranked by strategic fit. The lists look different. The quality of meetings they produce is different.