LinkedIn Prospect Research Prompt
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. The flagship research prompt. It takes a profile and whatever else you paste, and returns the four things that actually change a first message: what this person owns, what changed recently, what your product does about it, and why they might be the wrong person entirely. Everything it could not verify is listed under "unverified" rather than smuggled into the message as a fact.
You are a B2B research analyst preparing a seller for first contact. You are precise about the difference between what you read and what you inferred, because the seller will quote you to the prospect.
## Objective
{{OBJECTIVE}}
## Who I am
{{ROLE}} at {{COMPANY}}.
What we sell: {{PRODUCT}}
Proof I am allowed to cite: {{PROOF}}
## Who I am researching
Name: {{PROSPECT_NAME}}
Title: {{PROSPECT_ROLE}}
Company: {{PROSPECT_COMPANY}}
Profile and activity, pasted verbatim:
{{PROSPECT_PROFILE}}
[OPTIONAL: paste the company's LinkedIn About section, recent company posts, and any open roles they are advertising.]
## Method
1. Read before you infer. Build the facts section only from text I pasted.
2. Establish what this person is ACCOUNTABLE for, which is usually narrower than their title implies. A VP Finance at a 300-person company signs for software; a VP Finance at a 30,000-person company does not.
3. Find what changed. Rank changes by recency and by how much they disturb the status quo: a new job in the last 90 days, a reorganisation, a funding event, a hiring pattern, a product launch, a public complaint about a process.
4. Form ONE value hypothesis — a specific sentence about what my product would change for this person given what you found. Not a list of benefits. One sentence, falsifiable.
5. Actively look for the reason this is the wrong person or the wrong time, and say so. A disqualification is a better outcome than a polite no.
6. Rate your confidence in the hypothesis: high only if it rests on something explicitly stated in the pasted text.
## Constraints
- Never state a fact that is not in the pasted material. If you believe something is true but did not read it, it goes under "Unverified assumptions", not in the brief.
- Do not guess headcount, revenue, funding or tenure. Absent means absent.
- Do not flatter. "Impressive background" and "clearly a leader in the space" are noise and will be cut.
- The draft message must be under 300 characters and must contain exactly one specific, checkable reference to this person or company. Not two.
- Do not mention my product in the first message unless the value hypothesis is rated high confidence.
## Output format
### Who they are
Two lines: what they own, and what they are measured on.
### What changed
Up to three items, each with the evidence it came from and roughly when. Most recent first. If nothing changed, say "no observable change" — do not manufacture one.
### Value hypothesis
One sentence. Then: confidence (high / medium / low) and the single piece of evidence it rests on.
### Why this might be wrong
The strongest case that this person is not the buyer, or that now is the wrong time.
### Unverified assumptions
Everything you believe but did not read. Bullet list.
### Opening message
Under 300 characters. One specific reference. Ends with a question they can answer in one line.
### If they do not reply
One follow-up, sent no sooner than five business days later, that adds something rather than asking again.- You are about to send a connection request or an InMail to someone who matters enough to spend five minutes on.
- You have a list of 30 accounts and need a reason to contact each one that is not "I saw you work at".
- You are preparing for a call and want the three things you would have had to read four pages to find.
- You want an agent to do this at volume and need the instruction that makes the output consistent.
- Open the prospect’s profile. Copy the headline, the About section and the current role description — the whole thing, not a summary.
- Add anything they posted or commented on in the last two months. This is where the "what changed" section gets its best material.
- Paste the company’s About section and any open roles if the deal is worth the extra minute. Open roles are the highest-signal free data on LinkedIn.
- Run it, then read "Why this might be wrong" before anything else. About one prospect in five gets disqualified here, which is the point.
- Send the opening message as-is or not at all. Editing it until it sounds like your other messages usually removes the specific reference that made it work.
The output schema it demands
The single highest-leverage part of any prompt. Without a declared shape you get prose you then have to reformat; with one, you get sections you can act on — and you can tell at a glance when the model skipped something.
### Who they are ### What changed ### Value hypothesis (+ confidence + the evidence it rests on) ### Why this might be wrong ### Unverified assumptions ### Opening message (<300 characters, one checkable reference) ### If they do not reply
What it looks like filled in
Illustrative, not a promise — this is the shape of a good answer rather than a guarantee of one. The scenario is the same fictional company used across the whole toolkit.
Prospect: Dana Whitfield, Group Finance Director at Northbeam Logistics (~900 staff, 5 countries). Pasted: About section, current role, one post from last week about a five-week close after an acquisition. Seller: founder at Cadence.
Value hypothesis — Dana's close is long because the acquisition added a fifth entity to a process built for four, not because her team is slow; Cadence removes the manual inter-entity reconciliation that grows with each entity. Confidence: high — rests on her own post describing a five-week close "since the acquisition".
Why this might be wrong — She describes the problem as temporary ("since the acquisition"), which suggests she may be waiting for it to settle rather than shopping. If the acquisition closed more than two quarters ago, the window has probably passed.
Opening message — "Dana — your post on the five-week close after the acquisition. The bit I keep seeing: entity count, not team size, is what sets close length. Curious whether the fifth entity added days to reconciliation specifically, or across the board?" (243 characters)
Do not do these
Each one is a real failure mode, not a disclaimer.
- Pasting a link instead of the text. Most assistants cannot open LinkedIn — it blocks them — and a model that cannot read the page will cheerfully invent its contents.
- Running it on 200 prospects. At that volume you do not need research, you need a better filter; use the Lead List Qualifier first and research the survivors.
- Using the opening message with the reference removed because it felt too specific. The specificity is the entire mechanism.
Attach these once and stop retyping your business
This prompt works pasted into a blank chat. It works considerably better inside a Project with these attached, because then the ICP, the product and the limits are already in context and every answer inherits them.
ICP.md
Every prompt you run afterwards knows who you sell to, in terms LinkedIn can filter on.
Free downloadPRODUCT.md
Messages stop describing features and start describing what changes for the person receiving them.
Free downloadRESEARCH.md
Research that is consistent between people and between runs, with the evidence line drawn in the same place every time.
Free downloadThis is one step in a sequence
The output of the step before is the input to this one, and this one's output is the next one's input. That is the whole reason these are not independent pages.
Before you run it
Can ChatGPT or Claude read a LinkedIn profile from a URL?
Generally no. LinkedIn blocks assistant crawlers, so a model given a profile URL will either say it cannot access it or — worse, and common — produce a plausible profile it invented. Paste the text. Every prompt in this library is written to be fed pasted text for that reason.
Why does it refuse to mention the product in a low-confidence message?
Because a product mention converts a question into a pitch, and a pitch built on a weak hypothesis is the message that gets you marked as spam. When the evidence is thin the correct first move is a question you actually want the answer to.
Does this work for account research rather than one person?
Partly. It will handle a company page, but the "what they are measured on" section is the valuable part and it is personal. For account-level work, run this once per member of the buying committee and compare the three briefs — the disagreements between them are usually where the deal is.
The reference behind this prompt
Research is only worth five minutes if the message that follows costs something. On Premium Career an InMail costs $8 of your allowance; on Sales Navigator Core, $2.40.
The alerts that surface "what changed" without you checking — job changes, posts, company news — are a Sales Navigator feature, not a LinkedIn one.
Written to a fixed structure and run through Claude and ChatGPT to confirm each one returns the output format it declares. That is not the same as a measured reply-rate benchmark, and nothing here claims one — if you run these at volume and have numbers, we want them.
Take the whole folder instead
This prompt, the files it expects and the rest of the sequence, as a zip you can drop into a Claude Project, a Cursor workspace or an agent repo. Free, no account.