Cold LinkedIn Messages: The Three Rulebooks
Three separate rulebooks govern a cold LinkedIn message, and each one punishes something different.
Ethical LinkedIn outreach covers three separate things: what LinkedIn’s own rules forbid, what data protection law requires before you contact a stranger for business, and what only improves your reply rate. Sorting a guideline into one of the three tells you what ignoring it costs.
Each section below says which rulebook a guideline comes from, and the checklist near the end puts every line in one place. Automation shows up in all three, so it gets its own answer: accounts you own, one isolated session each, and limits enforced server-side before an action fires. On that stack, gtm-api.com reports 20,000+ accounts at under a 1% ban rate.
On this page
The three rulebooks behind any outreach guideline
Any guideline about LinkedIn outreach comes from one of three places: LinkedIn’s own User Agreement and Professional Community Policies, the data protection law that covers the person you are writing to, or somebody’s reply-rate data. Different people enforce each one, on different timescales, and ignoring one costs a different thing from ignoring the others.
That sorting is most of the work. Enforcement under the first rulebook is automatic and often silent, the second arrives through a regulator after somebody complains, and the third has nothing behind it except your own numbers, which is why it can be tested and dropped.
| Rulebook | Who enforces it | What ignoring it costs | A guideline from it |
|---|---|---|---|
| LinkedIn’s own rules User Agreement, Professional Community Policies |
LinkedIn, automatically, with no notice and no appeal at the lower tiers | Reduced reach, an invitation block, a restricted account, a permanent ban | Do not use bots or scrapers to reach the site |
| Data protection law GDPR, ePrivacy and PECR, US state laws |
A regulator, usually after somebody complains | Orders to delete, undertakings, fines that scale with turnover | Tell people where you got their details |
| Best practice Vendor benchmarks, sales folklore |
Nobody | A weaker reply rate, and only if the advice is right | Keep the first message short |
What LinkedIn’s rules actually forbid
LinkedIn’s rules are about access and identity. Nothing in them asks for a recipient’s permission before an invitation or a first message, and cold outreach itself is not a violation. What they prohibit is automated access, scraping, running an identity that is not yours, and taking member data off the platform for a competing service. Volume sits under a separate mechanism, the caps on LinkedIn limits.
The prohibitions sit in section 8.2 of the User Agreement, quoted word for word by the Ninth Circuit in hiQ Labs, Inc. v. LinkedIn Corp., No. 17-16783, decided 18 April 2022. Members agree not to:
- Automate access. “Use bots or other automated methods to access the Services.”
- Scrape or copy profiles. “Scrape or copy profiles and information of others through any means (including crawlers, browser plugins and add-ons, and any other technology or manual work).” The parenthesis is the part people miss: a browser extension and a hand-built export are named alongside a crawler.
- Feed a competing product. “Copy or use the information, content or data on LinkedIn in connection with a competitive service (as determined by LinkedIn).”
Two more rules matter as much for outreach and live outside 8.2. LinkedIn allows one profile per real person, so a persona account, a shared login or a bought profile breaks a rule before the first message goes anywhere. And the Professional Community Policies set out what enforcement looks like: “Depending on the severity of violation, we may limit the visibility of certain content, label it, or remove it entirely. Repeated or egregious offenses can result in account restriction.”
Etiquette re-enters through the report button. Recipients decide what counts as spam, and their decisions are an input LinkedIn acts on: invitations that get ignored or answered with “I don’t know this person”, and messages that get reported, feed the invitation restrictions LinkedIn hands out. The full trigger list and every published cap live on LinkedIn connection limits and LinkedIn limits. So a third-rulebook habit becomes a first-rulebook problem once enough people react badly.
What happens after a rule breaks belongs to other pages: the safety stack on safe LinkedIn automation, the appeal path on LinkedIn account restricted, and the silent reach penalty on LinkedIn shadow ban.
What the law requires when you contact a stranger for business
Cold B2B outreach is lawful in the EU, the UK and the United States. The law attaches conditions to it: a lawful basis for holding the person’s details, a duty to tell them you hold them, an obligation to stop when they object, and in the EU and UK a consent rule for some channels that plausibly reaches a LinkedIn message. Which of those bite depends on where your recipient is.
| Where the recipient is | Before the first message | After it |
|---|---|---|
| EU / EEA | A lawful basis for the record you hold, in practice legitimate interests under Article 6(1)(f), with the balancing exercise actually done | Tell them within one month (Article 14), and stop on request (Article 21) |
| UK | The same UK GDPR duties, plus the PECR question: consent or soft opt-in if the recipient counts as an individual subscriber | Identify yourself, give a working way to opt out, honour it |
| United States, federal | Nothing required before an email or a LinkedIn message | CAN-SPAM duties attach to email: accurate headers, a real postal address, opt-outs honoured within 10 business days |
| California | Notice at or before collection | Business contacts have held CCPA access, deletion and opt-out rights since 1 January 2023 |
Start with the basis. Recital 47 of the GDPR says in one line that “the processing of personal data for direct marketing purposes may be regarded as carried out for a legitimate interest”, which is why almost nobody collects consent for B2B prospecting. The same recital asks for “careful assessment including whether a data subject can reasonably expect” the processing, and that assessment is a test you can fail. In October 2024 the Irish Data Protection Commission fined LinkedIn Ireland 310 million euros over its advertising processing, finding the company “did not validly rely on Article 6(1)(f) GDPR (legitimate interests)” because its interests “were overridden by the interests and fundamental rights and freedoms of data subjects”.
The duty most outreach teams have never read is Article 14, which covers personal data you did not get from the person. It requires you to tell them, “within a reasonable period after obtaining the personal data, but at the latest within one month”. A first message that says where their details came from does that job in a clause.
Then the channel rule, and this is where a LinkedIn message stops resembling a work email. Article 13 of the ePrivacy Directive allows direct marketing by “electronic mail” only “in respect of subscribers who have given their prior consent”, and the directive defines electronic mail broadly as “any text, voice, sound or image message sent over a public communications network which can be stored in the network or in the recipient’s terminal equipment until it is collected by the recipient”. The UK regulator applies that definition to social platforms by name: the ICO’s key concepts guidance, updated 28 April 2026, lists “direct messaging on social media (ie private messages)” as electronic mail alongside email, SMS and in-app messages.
The escape hatch for B2B email is the subscriber test, and it travels badly to LinkedIn. The ICO’s business-to-business guidance of 22 May 2025 says “the PECR rule on direct marketing by electronic mail does not apply to corporate subscribers”, where a corporate subscriber is a limited company, an LLP or a Scottish partnership, while “people, sole traders, ordinary partnerships” are individual subscribers who need consent or a soft opt-in. A work email address belongs to the company. A LinkedIn account belongs to the person, and it survives the job. No regulator has published a decision applying the subscriber test to a LinkedIn message, so treat it as the open question it is.
France’s regulator states the B2B position in the plainest terms available. The CNIL page on electronic prospecting, updated 10 June 2026, holds that prospecting a professional needs no prior consent as long as the message is “en rapport avec la profession de la personne démarchée”, relevant to that person’s job, the person was told their details could be used this way, and they are “en mesure de s’y opposer”, able to object. Relevance is doing legal work in that sentence, which matters for the section below.
US federal law is narrower than its reputation. CAN-SPAM regulates a “commercial electronic mail message”, and 15 U.S.C. 7702 defines an electronic mail message as “a message sent to a unique electronic mail address”, built from “a unique user name or mailbox” and “a reference to an Internet domain”. A LinkedIn thread has no such address, so the FTC’s requirements, accurate headers, a valid physical postal address, an opt-out honoured within 10 business days and penalties up to $53,088 per offending email, follow your email programme and stop at the edge of the platform. State privacy law is what reaches a US prospect: California’s exemption for business-to-business contact data expired on 31 December 2022, so a prospect there has access and deletion rights over whatever your CRM holds.
Guidelines that are neither required nor effective
Some outreach advice gets repeated until it is treated as a rule. The ones below fail both tests: they appear in no rulebook, and the published data does not carry them either. Dropping them costs nothing and frees up the attention the first two rulebooks need.
Always personalise the connection request. LinkedIn requires no note, and it caps free accounts at a handful of personalised notes a month anyway, a limit that lives on LinkedIn connection limits with the number and the source. The datasets that measure the note disagree with each other about which direction it moves:
| Dataset | Sample | What it reports |
|---|---|---|
| SmartReach, State of LinkedIn Outreach, 7 August 2026 | 500,000+ connection requests, January to June 2026 | Acceptance 24% with a note against 28% without. Replies per accepted connection 22% with a note against 14% without |
| Botdog, 27 November 2025 | 16,492 invitations | Blank requests accepted at a higher rate, on the reading that they feel less like a pitch |
| Expandi, quoted in Overloop’s benchmark round-up, 11 June 2026 | 13.2 million connection requests, May 2025 to April 2026 | Around 45% acceptance for personalised requests against around 15% for generic ones |
Every number there was published by a company selling the sending tool that produced it, and they point in opposite directions on the same question, which is a good reason to treat the note as a test you run on your own list. The SmartReach split suggests what a note actually buys: much the same acceptance, and a warmer thread once the connection lands.
Send at the optimal time. Reachium measured 196,696 connection requests sent between 8 January 2025 and 27 July 2026 and found acceptance by day of week spanning 26.38% to 28.24%, a spread of 1.86 points, with weekend sends landing within 0.27 points of weekday sends. Their write-up also names the flaw in every hourly chart you have seen: “Hour-of-day charts measure who sends, not when works.”
Paste an unsubscribe line into the message. This one is imported from email compliance, where it is mandatory. CAN-SPAM does not reach a LinkedIn thread, and the EU and UK duty is to honour an objection when somebody makes one, which no footer creates or discharges. It spends a line of a short message and buys nothing legally.
The practices that do move replies
The practices with evidence behind them all make a message easier to answer: relevance to the person’s job, a short first message, a stated reason you are writing to them specifically, and one follow-up. Two of them are also legal duties in the EU and the UK, which is the overlap worth designing around.
- Relevance to the person’s role. The CNIL makes it the test that keeps B2B prospecting outside the consent rule in France, and it is also what separates a message somebody answers from one somebody reports.
- Say where their details came from. Article 14 requires the disclosure within a month, and putting it in the first message answers the question every cold recipient asks first.
- Keep the first message short. Overloop’s 2026 benchmark round-up, quoting LinkedIn’s own InMail data, puts messages under 400 characters at 22% higher response than average, with messages over 1,200 characters landing 11% below it.
- One follow-up, then stop. A second nudge is normal. A fourth is how a thread turns into a report, and reports feed the platform triggers above.
- Honour a no on the spot. Article 21(3) is blunt about it: “Where the data subject objects to processing for direct marketing purposes, the personal data shall no longer be processed for such purposes.” That covers the CRM record as well as the thread.
- Send a volume you can defend. Pace is where good practice and account safety meet. The published ceilings and the safe daily defaults are on LinkedIn limits, and spreading volume across several senders is covered on scaling LinkedIn outreach.
The pre-send checklist
One list, each line tagged with the rulebook it comes from. The Rule lines put the sending account at risk. The Law lines put the company at risk. The Practice lines put the reply rate at risk, and you are free to test them against your own numbers.
- The account belongs to a real person, and that person is you or your team. (Rule) One profile per person, no bought or invented identities, no shared logins between operators.
- Nothing in the list was scraped from profiles. (Rule and Law) Section 8.2 names crawlers, browser plugins and manual work in one breath, and a scraped record still needs a lawful basis and an Article 14 notice. Whether the collection itself is lawful is a separate question with three separate answers, worked through on is scraping LinkedIn legal.
- Anything automated runs on an account you connected yourself. (Rule) One isolated session each, human pacing, limits checked before the action fires. Mechanics on safe LinkedIn automation, and the case against page-parsing tools on LinkedIn scraping vs a safe API.
- You can name the lawful basis for the CRM record. (Law) For B2B prospecting it is legitimate interests, and it holds only with the balancing exercise done and written down.
- The first message says where you got their details. (Law and Practice) Article 14 within a month, and the recipient’s first question answered in the same clause.
- An objection stops everything within a day. (Law) The sequence, the CRM record, the retargeting list. Article 21 makes the right absolute for direct marketing.
- The message names why this person, in their words. (Practice, and the CNIL test) Role relevance carries legal weight and reply rate in one line.
- Volume is inside the published caps, with headroom. (Rule and Practice) Ceilings and safe defaults are on LinkedIn limits.
Where automation sits in all three
Automation changes none of the three rulebooks. It changes the speed at which an account can break the first one, and the volume of personal data the second one applies to. The useful controls are the ones enforced before an action fires, while a decision can still be stopped.
GTM API takes the LinkedIn side of that. Accounts are ones you connect and own, each running in its own isolated cloud browser session on a dedicated proxy, with a daily limit checked server-side before every outward action and an audit log of what was sent to whom. Its LinkedIn side exposes 160+ typed tools across 11 toolsets, and outward actions can require a preview and a confirmation before anything leaves the account. LinkedIn is the only channel it sends on today. On that stack, gtm-api.com reports 20,000+ accounts at under a 1% ban rate.
None of that answers the second rulebook for you. A server can hold your sending inside a cap and log what happened, and a person still has to decide the basis for the list, write the line that says where the details came from, and take a no as final.
Frequently Asked Questions
Do you need someone’s consent before messaging them on LinkedIn?
Not under LinkedIn’s own rules, which say nothing about permission before an invitation or a first message. Consent is a data protection question, and the answer is regional. In the EU and the UK, direct marketing by electronic mail needs consent or a soft opt-in where the recipient is an individual subscriber, and the ICO counts private messages on social media as electronic mail. Legitimate interests covers most B2B prospecting, provided the person is told and can object.
Does GDPR apply if I only message people inside LinkedIn?
It applies as soon as you hold their details anywhere of your own. Reading a public profile is LinkedIn’s processing under LinkedIn’s notices. Copying the name, role and company into a CRM, a sequence or a spreadsheet makes you the controller of that record, which brings a lawful basis, the Article 14 duty to tell the person within one month, and the absolute right to object to direct marketing under Article 21.
Does CAN-SPAM cover LinkedIn messages?
No. CAN-SPAM regulates a commercial electronic mail message, and 15 U.S.C. 7702 defines an electronic mail message as “a message sent to a unique electronic mail address”, meaning a mailbox name plus an internet domain. A LinkedIn thread has no such address. The Act’s requirements, including the valid physical postal address and the opt-out honoured within 10 business days, follow your email programme and stop at the edge of the platform.
Can I import a purchased contact list and message those people on LinkedIn?
Both of the first two rulebooks make this expensive. If the list was built by scraping profiles, section 8.2 of the User Agreement was broken before you sent anything, and it names manual work and browser plugins alongside crawlers. Under GDPR a bought list still needs a lawful basis you can state, and the people on it have to be told you hold their data within a month. Consent the seller collected only helps if it named you.
Should I add a note to a LinkedIn connection request?
It is a test to run on your own list, and it is a requirement nowhere. Vendor datasets published across 2025 and 2026 contradict each other: one platform’s 500,000+ requests show acceptance of 24% with a note against 28% without, while a 13.2 million request dataset from another vendor reports roughly three times the acceptance for personalised requests. The finding worth acting on is what a note does after acceptance, where the same 500,000-request dataset reports replies of 22% with a note against 14% without.
When does LinkedIn outreach become spam?
Recipients decide, and on LinkedIn their decision reaches the sender: ignored invitations, “I don’t know this person” answers and reported messages all feed the triggers behind an invitation restriction. The practical test is whether the person can tell in one line why they were picked. Volume without that line is what gets reported, and the report is what turns manners into account access.
Sources & Further Reading
- hiQ Labs, Inc. v. LinkedIn Corp., No. 17-16783 (9th Cir., 18 April 2022), footnote 6, which reproduces the section 8.2 prohibitions quoted above as the agreement stood in that case. The live text is in LinkedIn’s User Agreement.
- LinkedIn, Professional Community Policies: “Depending on the severity of violation, we may limit the visibility of certain content, label it, or remove it entirely. Repeated or egregious offenses can result in account restriction.” Read 29 August 2026.
- Regulation (EU) 2016/679, GDPR: Recital 47 on direct marketing as a legitimate interest, Article 14(3)(a) on the one-month deadline, Article 21(2) and 21(3) on objecting to direct marketing.
- Directive 2002/58/EC, ePrivacy: Article 13(1) on prior consent for direct marketing by electronic mail, Article 2 for the definition this page relies on.
- ICO, key concepts for direct marketing using electronic mail (updated 28 April 2026), which lists “direct messaging on social media (ie private messages)”, and business-to-business marketing (22 May 2025) for the corporate and individual subscriber split.
- CNIL, la prospection commerciale par courrier électronique (updated 10 June 2026), the B2B position quoted above.
- Data Protection Commission (Ireland), press release of 24 October 2024 on the 310 million euro fine and the failed legitimate-interests reliance.
- FTC, CAN-SPAM Act: A Compliance Guide for Business, and the statutory definitions at 15 U.S.C. 7702.
- California Attorney General, CCPA FAQ: “The exemptions for employment-related personal information and personal information reflecting business-to-business transactions described in Civil Code Sec. 1798.145(m)-(n) expired on December 31, 2022.”
- Connection-note and timing figures, all vendor-published and fetched 4 September 2026: SmartReach, State of LinkedIn Outreach (7 August 2026), Botdog (27 November 2025), Overloop’s benchmark round-up (11 June 2026, quoting Expandi), and Reachium (27 August 2026). Each vendor sells a sending tool, so these are corroboration between competitors, not authority.
- Related: Is scraping LinkedIn legal · Safe LinkedIn automation · LinkedIn limits · LinkedIn connection limits · LinkedIn account restricted · LinkedIn shadow ban · LinkedIn scraping vs a safe API · Scaling LinkedIn outreach
Messages that survive all three rulebooks.
Owned accounts, one isolated session each, server-side limits before every send, and an audit log of what went to whom. On that stack, gtm-api.com reports 20,000+ accounts at under a 1% ban rate. Free plan, then from $10 per account at volume.
Last updated: September 2026 · Legal texts and regulator guidance checked against the live pages on 4 September 2026. The subscriber question for LinkedIn messages is open, and this page says so where it comes up.