By KYN AI Advisory Team — AI implementation specialists, Singapore
LinkedIn prohibits third-party software that scrapes its site or automates activity on it, and its help page on prohibited software says this covers crawlers, bots and browser extensions. The risks fall on the account that runs the tool: restriction or closure, and tools that stop working without notice. For an SME, the practical answer is to build prospecting on sources you are allowed to use, and to ask any AI vendor exactly how it gets its LinkedIn data.
What does LinkedIn actually prohibit?
LinkedIn's help page states that it does not permit third-party software that scrapes or modifies its site, or automates activity. It names Section 8.2 of the User Agreement. Prohibited activity includes:
- using automation to add contacts, send messages, or like and comment on posts;
- copying or scraping profiles and other data;
- overriding security features or access controls.
The page also states that members who use such tools risk having accounts restricted or shut down, and that LinkedIn keeps improving its technical defences. That is LinkedIn's position, stated in its own words. Read the page itself before you decide.
Who carries the risk when a tool breaks the rules?
The person whose account is used. If a vendor's agent logs in as your salesperson and browses 500 profiles an hour, the restriction lands on that salesperson's profile, which is also a business asset: their network, their connections and their history.
The vendor usually carries less. Its terms commonly place compliance on the customer. Read the contract for a clause that says so.
There is a second risk: the tool stops. Platforms change page structure and add defences. A workflow built on scraped data can fail overnight, and no one is accountable for the gap.
Is scraped data also a privacy problem?
It can be. Profiles hold personal data. Collecting it for marketing raises questions under data protection law in each country where the person lives. In Singapore, the PDPC's Do Not Call Registry guidance applies to marketing messages sent to Singapore numbers. Hong Kong, Indonesia and Malaysia each have a regime of their own. Public visibility does not by itself make a use lawful.
This is not legal advice. Ask a lawyer about your specific use, particularly if you contact people in several countries.
Which sources are safer for prospecting?
Build the account list and the research from sources that are yours or that publish for reuse.
| Source | What it gives | Notes | |---|---|---| | Your own CRM and inbox | Past customers, replies, enquiries | You hold the relationship | | Your website analytics | Which companies visit which pages | First-party; check consent notices | | Company registries | Legal name, registered number, directors, address | In Singapore, ACRA records; each registry has its own terms | | Company websites | Team pages, news, services | Public pages; respect robots.txt | | Licensed data providers | Contact data with a stated source | Ask for the source and the opt-out process | | Manual LinkedIn use | A salesperson reading a profile | Normal use of the product | | LinkedIn's official products | Sales Navigator and approved partner integrations | Check LinkedIn's terms for what each allows |
A person using LinkedIn to read a profile before a call is ordinary use. An agent harvesting profiles in bulk is the thing the rules address.
How do you review an AI prospecting vendor?
Ask five questions in writing.
- Where does each field come from? Ask for a source per data field: registry, website, licensed provider, LinkedIn.
- Does the tool log in to LinkedIn as a user, or use a licensed interface?
- Who is responsible if an account is restricted? Look for it in the contract.
- How are opt-outs honoured? A person who asks to be removed should disappear from every list and every channel.
- What happens when a source changes? You want a defined fallback, not silence.
If a vendor cannot answer the first two, treat it as scraping.
Also see our notes on asking an AI vendor about security.
What does a lower-risk prospecting workflow look like?
- A person defines the target: country, entity type, size band, sector.
- The agent builds a candidate list from registries and company websites.
- For each account, the agent writes one checkable fact and cites the page it came from.
- A person reviews the list and the draft first line.
- A person looks up the buyer on LinkedIn by hand if needed, and sends a connection request or message themselves.
- The agent logs the result, including opt-outs, to the CRM.
This keeps automation on the research and the log, where a mistake is cheap, and keeps people on the actions that carry your name. It matches the pattern in what an AI sales agent can and cannot do.
What goes into a vendor review file?
Keep one page per vendor. It takes about an hour to fill and protects you if a tool is later restricted. An illustrative layout:
| Question | Answer to record | Evidence | |---|---|---| | Data fields collected | List each field | Vendor documentation | | Source per field | Registry, website, licensed provider, LinkedIn | Vendor reply in writing | | LinkedIn access method | Licensed interface, or user login | Vendor reply in writing | | Account at risk | Whose login is used | Contract clause | | Opt-out handling | How a removal reaches every list | Test: submit one | | Sub-processors and location | Where data is stored | Privacy terms | | Exit | What you keep if you stop | Contract clause |
Run one test before signing. Submit a removal request for a test contact, then check that the contact disappears from every list and channel. A vendor that cannot do this will not handle a real opt-out.
Review the file each year, or when the vendor changes its terms or its method. Keep the previous version, so you can show what you knew and when. If a vendor refuses to fill the file, record that refusal with the date. It is itself an answer, and it belongs in the decision.
Assign one named owner for the file. When a salesperson asks to try a new prospecting tool, the owner opens a new page before the trial starts, not after the first account warning.
Where KYN fits
KYN builds prospecting on registry records, company websites, your CRM and your own enquiries. Each first line cites a source a reviewer can open. Agents use their own restricted accounts, never a salesperson's login, and every action is logged with the actor. Outbound messages wait for a named person. For an insurance brokerage, outreach with one checkable first line costs about S$0.02 per email.
The first 14 days are a review of where your prospect data comes from today, with a source-by-source risk note. A full build ships in under 90 days. See sales systems and security.
Frequently asked questions
Can I use an AI tool to write LinkedIn messages if I send them myself?
Drafting is different from automated sending. LinkedIn's page targets tools that scrape or automate activity. A draft you read, edit and send yourself is your own message. Check the current User Agreement for the detail.
Is public LinkedIn data free to use?
Visible does not mean free to collect. LinkedIn's terms restrict scraping, and data protection law may apply to the personal data in a profile. Treat public as a starting point for a question, not a licence.
What if a vendor says its tool is compliant?
Ask how, in writing, and for the source of each data field. A general claim is not an answer. If the method involves logging in as a user and automating clicks, LinkedIn's page describes that as prohibited.
Checked 30 October 2026.
Curious whether Marketing agents fits your business? Talk to KYN on WhatsApp — no forms, just a conversation.
