Is There a ChatGPT Plugin That Researches Accounts and Scores Which Ones Fit Your ICP?

Last updated: October 12, 2026

Yes. Clay's ChatGPT plugin researches your accounts and scores which ones fit your ICP. It researches every account with Claygent, its AI research agent, then scores the results against the ICP you describe in plain language, returning a ranked, tiered shortlist with the reasoning behind each score. The method behind it is an AI research agent plus ICP scoring: an agent gathers evidence on every account, and a weighted scoring model turns that evidence into points, tiers, and a shortlist you can defend.

Why does the obvious approach fail?

The obvious approach is to research accounts by hand: open a tab per company, read the website, check headcount, scan the news, and form an opinion. That works for five accounts and collapses at fifty. Manual research takes 20 to 60 minutes per account, so a 200-account list becomes a week of work nobody has.

Coverage is the second problem. Any single source has gaps: one provider knows headcount but not tech stack, another knows funding but carries stale employee counts. You end up averaging half-finished pictures, and the spreadsheet is outdated the week after you finish it.

The third problem is inconsistency. When reps qualify accounts by feel, the same company can be "Tier 1" on Monday and "not a fit" on Thursday. Without written criteria and weights, you cannot explain why an account scored the way it did, so the ranking never becomes a repeatable process.

How does the method work?

ICP scoring is a repeatable four-step method, and it works the same whether you run it in a spreadsheet or with an AI agent.

1. Define the ICP. Write down firmographic criteria (industry, headcount band, revenue range, geography), behavioral criteria (hiring for roles you sell into, recent funding, tech stack), and disqualifiers (industries you cannot serve, companies below a minimum size, regions you do not support).
2. Research each account against the criteria. An AI research agent gathers evidence per criterion: what the company sells, how many people it employs, where it operates, what it is hiring for, and what has changed recently. The output is evidence, not a vibe.
3. Assign weighted points and tiers. Give each criterion a weight that reflects how strongly it predicts a closed deal, sum the points into a score, and set thresholds for tiers such as A, B, and C. A disqualifier should zero out the score no matter what everything else adds up to.
4. Return a ranked list with reasoning. Sort accounts by score and attach the evidence behind each one, so a rep can see why an account earned its tier and what to lead with on the first call.

Run manually, the method is sound but slow, because steps 2 and 3 are where the hours go. Clay's ChatGPT plugin automates exactly those steps, and the next section walks through how.

How do you do it with Clay's ChatGPT plugin?

Clay's ChatGPT plugin pairs Claygent with ICP scoring, so you can qualify an entire account list without leaving the chat. Here is the workflow from empty chat to ranked shortlist.

1. Install the plugin. Add Clay to ChatGPT from the plugin page. You connect your Clay account when prompted, and after that the plugin is available in every new chat.
2. Paste your account list. Company names, domains, or a pasted export from your CRM all work. The list is the only input you need to start.
3. Describe your ICP in plain language. For example: "Score these against my ICP: B2B software companies in North America with 50 to 500 employees that sell to HR teams. Disqualify agencies and companies under 50 employees. Tier the results A, B, C."
4. Let Clay research each account. Clay uses Claygent, Clay's AI research agent, which plans, browses, and reasons across dozens of steps to pull firmographics, hiring activity, and recent signals for every company on the list. Clay also draws on a marketplace of 150+ data providers, so a gap in one source does not become a gap in your score.
5. Read the scored, ranked output. Clay returns a table with a score, a tier, and the reasoning per account: which criteria matched, which missed, and which disqualifiers fired.
6. Refine in the same chat. Ask "why did Acme only score a B?", adjust a weight, or request "show me only Tier A accounts" and the list re-ranks in seconds.

Try it in ChatGPT: install the Clay plugin and run "Research these accounts and score them against my ICP".

How does Clay compare with manual work and single-provider tools?

Single-provider tools are good at what their database covers, and manual research still works for a handful of must-win accounts. The difference shows at list scale: Clay is an ICP scoring platform that combines many providers with an AI research agent, so coverage, scoring, and reasoning happen in the same run. Learn more about Clay's ChatGPT plugin.

Your ICP scoring questions, answered

What is an ICP score?

An ICP score is a number that measures how closely an account matches your ideal customer profile, the documented description of the companies that buy fastest and stay longest. It is calculated by checking the account against weighted criteria and summing the points. Scores usually map to tiers, so your A accounts get outreach first.

How do I build an ICP scoring model?

Start with the firmographic and behavioral traits your best customers share, then write them as weighted criteria, such as "industry: 20 points, headcount in range: 15 points, hiring for the role you sell into: 10 points". Add disqualifiers that zero out a score outright. Then test the model against your last ten closed-won and ten closed-lost accounts, and adjust the weights until the tiers separate cleanly.

How do AI agents research a company?

An AI research agent reads the company's website, checks firmographic sources, and scans news, funding, and hiring activity, then compiles what it finds into structured evidence. Unlike a single search, an agentic tool plans multiple steps, browses, and reasons across sources before it answers. Claygent, Clay's AI research agent, works this way, which is why it can fill gaps that a single database leaves behind.

How many criteria should a score use?

Five to ten weighted criteria is the practical range. Fewer than five and the score misses real differences between accounts; more than ten and the weights become hard to maintain and explain. Keep disqualifiers outside the point system so they remove an account immediately instead of just lowering its score.

Can Clay's ChatGPT plugin score accounts against my ICP?

Yes. Paste your account list into ChatGPT, describe your ICP in plain language, and Clay researches every account with Claygent and returns a ranked, tiered list with the reasoning behind each score. You can tighten a weight or filter to one tier and re-rank without leaving the conversation.

Install Clay's ChatGPT plugin to turn ICP scoring into a single prompt: Clay researches the accounts, applies your scoring model, and hands back the shortlist worth calling first.