AI for LinkedIn prospecting
Prospecting is the part where AI actually earns its place, because it is judgement applied to text at volume — and it is also where the published prompts do the most damage, by producing target lists five times larger than any LinkedIn account can work.
Build the ICP first. It is the input to everything else, and a search built without one is a search you will rebuild in a fortnight.
Researching everybody. Research is worth five minutes on a prospect you have already decided is worth contacting, and nothing on one you have not. Qualify first, research second.
4 for this work
- Prompt
ICP Builder
An ICP written in the vocabulary LinkedIn search understands, derived from deals you actually closed rather than from who you wish would buy.
- Prompt
LinkedIn Prospect Research
A one-page research brief on a single prospect that ends in a usable opening line, and that separates what it read from what it inferred.
- Prompt
Lead List Qualifier
Your list, split into three tiers with a one-line reason per row, sized to what a LinkedIn account can actually contact in a week.
- Prompt
Buying Signal Analyzer
A ranked signal list for your specific product, with the LinkedIn surface each one appears on and the window it stays actionable for.
Attach these once instead of retyping
A prompt is consumed once; a context file is read on every turn for months. These are the ones this work depends on.
- Context file
ICP.md
Every prompt you run afterwards knows who you sell to, in terms LinkedIn can filter on.
- Context file
COMPANY.md
The assistant stops describing your company the way its training data describes companies in your category.
- Context file
CLAUDE.md
An agent that reads the right files in the right order and knows what it is not allowed to do.
- Context file
AGENTS.md
Three narrow agents with defined handoffs instead of one conversation that drifts.
- Context file
RESEARCH.md
Research that is consistent between people and between runs, with the evidence line drawn in the same place every time.
These are steps in a longer job
From ICP to booked meeting runs 8 steps end to end, and several of them are on this page. Each step names what it hands the next one, which is the part a list of prompts cannot show.
LinkedIn’s own contribution to prospecting is a search box you can type a sentence into
It is free, it is on every account, and it is the only piece of prospecting AI LinkedIn gives away. The paid layer above it summarises a lead you have already found rather than finding you a better one — which is worth knowing before a tool bill is justified by it.
| What LinkedIn calls it | What it does | Availability | What unlocks it | What that costs |
|---|---|---|---|---|
| Search using keywords or natural, conversational language | You type the question the way you would say it instead of assembling filters, and LinkedIn returns results for it. | Available | Every account, free or paid | No charge on any plan |
| Filters in AI-powered people search | A filters layer over AI-powered people search. What the filters are is not public. | Gradual rollout | LinkedIn does not say | LinkedIn publishes neither the plan nor a price |
| Lead IQ | A generated summary of an individual lead, drawn from what they have posted and published. | Available | Sales Navigator Advanced or Sales Navigator Advanced Plus | $159.99 a month, read from LinkedIn on 12 September 2026 |
Feature names are LinkedIn's own, read from its help centre and product pages on 20 September 2026; the prices come from this site's plan records and carry their own read dates per product family. Availability is LinkedIn's word, not a measurement — a gradual rollout means some accounts have it and LinkedIn does not publish which. The full records, with the ceilings and what LinkedIn declines to disclose, are on the LinkedIn AI features page.
The research tools that read LinkedIn without your account being the thing that reads it
This is the one group where the route genuinely changes what you can buy. A data vendor sells a copy of the profile and never touches your login; an extension reads the page you already have open; a vendor cloud opens the page as you. The first two cannot get an account restricted, because no account of yours is used.
Everything on LinkedIn is outside them. Covering LinkedIn as well means buying, paying for and running a second product, and moving data between the two.
No account of yours can be restricted, because no account of yours is used. What it sells is a copy, as current as its last refresh — and nothing behind your login: not your network, not your inbox, not your saved lists.
It reaches whatever the page in front of you shows, and nothing while the browser is closed. LinkedIn’s Prohibited Software and Extensions policy names extensions that automate the site, and enforcement lands on the member account.
The widest reach of any class and the highest exposure. It runs without you, the credential sits with somebody else, and every action is attributable to your account — which is also the thing LinkedIn can restrict.
Unknown, which is the point. Nothing in this directory treats an unread claim as a capability, and nothing treats an unanswered question as a reassurance.
| Tool | How it reaches LinkedIn | What it does with your account | Account exposure |
|---|---|---|---|
| Landbase | No LinkedIn | No account used | None |
| Clay | Data vendor | No account used | None |
| Apollo.io | Chrome extension | Extracts automatically | Medium |
| PhantomBuster | Vendor cloud | Sends under your name | High |
| Common Room | Not established | No account used | Not established |
One tool per route, alphabetically first within each, so nothing here reads as a pick — and exposure is derived from the route and whether the tool sends, not typed next to the vendor it flatters. All 14 in this group, filterable, are in the LinkedIn Research Tools directory; every record was read from the vendor's own pages on 20 September 2026.
If you want the research fetched rather than pasted, that is what an MCP server is
A prompt on this page assumes you paste the profile in. An MCP server hands the assistant a tool that fetches it instead. 4 of the 13 in the directory need no LinkedIn account at all, which is the only arrangement that takes account exposure off the table.
A hosted, paid server that returns structured LinkedIn data — profiles, people search, companies, posts, jobs, the Ad Library — plus a dozen other sources, with server-side filtering and export. No LinkedIn login goes anywhere near it, and it cannot write to LinkedIn at all.
A general web-access server with five LinkedIn extractors: hand it a profile, company, job, post or people-search URL and it returns structured JSON. It holds no LinkedIn credential, so there is no account to restrict.
Research and record-keeping for manual outreach: it briefs you on a person or a company, mines a post's comment thread, and keeps a local ledger of what you sent and what replied. It has no send path at all.
Reads public LinkedIn company and person pages, and a company's public posts when LinkedIn serves them signed out. It never logs in, so it puts no account at risk — and cannot see anything behind the login.
4 of the 10 servers filed under profile and company data, ordered by account exposure, lowest first. That is an ordering, not a recommendation: LinkedIn publishes no MCP server of its own and its User Agreement prohibits automated access by any means, so the directory puts that clause beside each record rather than ranking them. Read 20 September 2026.
Three things no prompt on this page can do
Stated here rather than left for a reader to discover, because the gap between what a prompt does and what people expect it to do is where the time goes.
LinkedIn’s robots.txt blocks every major assistant crawler outright, so a model handed a profile URL will either refuse or invent the contents of the page and present them confidently. Every prompt here is written to work from text you paste in. What LinkedIn does and does not open to AI.
Nothing on LinkedIn’s self-serve permission list returns another member — not people search, not a profile. The prompt produces the filter set and the Boolean; the last three minutes are yours, in the LinkedIn interface. The 13 servers that try, and what each hands LinkedIn.
Regular LinkedIn search stops at 1,000 results and Sales Navigator stops at 2,500, per search rather than per day. A qualification prompt that hands you 5,000 names has produced a list the platform will not show you. Every ceiling, sourced and dated.
The questions people arrive with, answered directly
Can ChatGPT or Claude find prospects on LinkedIn for me?
No. LinkedIn blocks assistant crawlers in robots.txt and exposes no people-search API to a self-serve developer app, so an assistant given a LinkedIn URL either refuses or invents the page. What it can do is everything either side of the search: define the ICP, write the filter set and the Boolean, and score the list you export. The search itself stays a human clicking filters.
What is the best AI prompt for LinkedIn prospecting?
The one that knows what you can do afterwards. A prompt that returns 500 qualified accounts has produced 5 weeks of sending at roughly 100 invitations a week, and the bottom of that list is stale before you reach it. Start with the ICP Builder, then the search, then the qualifier — in that order, because each consumes the previous one’s output.
Do I need Sales Navigator to use these prompts?
No. Every prompt here works against regular LinkedIn search, which caps at 1,000 results per search. Sales Navigator raises that to 2,500 and adds the filters the search prompts mark as tier-gated; Core is $119.99 a month, read from LinkedIn on 12 September 2026. The prompts flag which filters a tier does not have rather than assuming you are on Advanced.
Is there an AI tool that does LinkedIn prospecting end to end?
Several claim to, and 9 of the 42 tools in the directory have no LinkedIn connection at all — their own documentation says they cover email and the phone. Of the ones that do reach LinkedIn, 18 act under your name, which is what LinkedIn’s enforcement is built to catch. The route matters more than the feature list, and it is the first thing the directory sorts on.
Everything for prospecting, in one folder
The Starter Pack bundles the twelve context files and the twelve most-used prompts as a zip — including the ones on this page. Free, and the ungated link sits under the form.