LinkedIn, recruitment, sales and AI
Research date: 4 October 2026
Edition: First baseline report
Method: Review of LinkedIn’s public documentation. No signed-in business account was inspected and no controls were tested in a live account.
Summary
LinkedIn’s AI supports visibility, writing, recruitment and sales research. Some recruitment permissions and member-data training preferences are documented as enabled by default.
- By default: Background personalisation, security and anti-abuse processing operate within the service. The member content-model training preference is documented as on unless switched off. For eligible recruitment users, Hiring Assistant, applicant evaluation and permission for automatic outreach have documented defaults; permission for outreach does not mean every project sends messages automatically.
- Beyond the default: Drafting posts, profiles and messages; candidate discovery, matching, summaries and outreach; Sales Navigator research and AI-assisted advertising.
- Outside LinkedIn: Sales Navigator can use linked websites, news and public company filings. CRM Sync can import accounts, contacts and leads and write messages and notes back to the CRM.
- Business exposure and learning: Recruitment AI can process resumes, screening answers and recruiter notes. LinkedIn’s reviewed Sales Navigator guidance says CRM Sync data is not processed by its generative features, although the integration uses it for other purposes. Member training, Sales Navigator protections and feedback submissions have different terms.
The main exposure: AI can influence who gets considered for a job and who receives sales messages, while CRM integration adds access and write-back beyond LinkedIn.
Business finding
LinkedIn qualifies for this research because a business can encounter AI through its professional presence, recruitment and sales work without buying a separate general-purpose AI service. The important distinction is between using AI, supplying information that AI processes, and allowing information to train AI models. These require different decisions.
LinkedIn provides a member setting for certain AI training and product-specific controls for recruitment automation. This review did not identify one switch that turns off every form of AI across LinkedIn. A company Page, an employee’s personal account and a purchased Recruiter or Sales Navigator seat should be assessed separately.
Where is AI operating?
LinkedIn describes AI-supported services including personalisation, content creation, security and anti-abuse functions. Its writing tools can help create posts, messages and profile content. These visible assistants sit alongside less visible processing that shapes the service. [1][2]
Hiring agents can find candidates, match qualifications, summarise applicants and contact people. Inputs include profile information, job-seeking preferences, available resumes and screening responses. Recruiters can add notes and other candidate information. The benefit is help with finding and reviewing candidates; the business still needs to understand the basis of the recommendations. [3]
Sales Navigator’s generative AI can process product information, buyer descriptions, public profiles and posts, linked websites, news and public company filings. It supports summaries and sales outreach. [4]
Advertising is another area to assess. LinkedIn’s guidance describes AI-powered Accelerate advertising alongside Classic options, although campaign interfaces and capabilities are changing. Choosing a campaign format should not be treated as proof that all automated advertising processing has stopped. [11]
For a small business, start with the activities actually undertaken: publishing, prospecting, recruiting, advertising and connecting customer systems. A business using LinkedIn only to publish posts has a different exposure from one connecting its sales database and using recruitment automation.
Who controls it?
Individual members control the training preference for their accounts. The training guide describes it as on by default unless switched off. This is a member choice; it should not be assumed that a company Page administrator can make the choice for every employee. [2]
Recruiter has additional business controls. Project owners can remove Hiring Assistant from projects they own. Designated administrators can control applicant evaluation and automatic outreach across users on their company contract. Access also depends on the purchased product and assigned permissions. [6][7][8]
For Sales Navigator enterprise seats, LinkedIn says the purchaser can request access to retained feature information and action on that information. This is different from an employee merely closing a feature session. [4]
The business should name the person responsible for each service and identify who can change its settings. An agency running advertising or recruitment needs to provide the business with the actual configuration, rather than simply saying that it uses LinkedIn.
What can the business switch off?
Member training preference
Open Settings & Privacy → Data privacy → Data for Generative AI Improvement and turn the training setting off. LinkedIn links directly to this preference from its training guidance. Save the choice and reopen it to confirm its state. [2][10]
This affects future covered training. It does not undo past training, disable features or cover other model purposes such as personalisation and security. [2]
Hiring Assistant in a project
In Recruiter, open Projects → the relevant project → Project settings → Workflow settings → Hiring Assistant automation → Remove from project. LinkedIn says this stops new sourcing, evaluation and assistance with messaging in that project. It remains active in other projects. Previously sourced candidates, summaries and past activity remain visible. [6]
The same guide says Hiring Assistant is enabled by default when eligible users create a project or post a job manually. Check new projects as well as existing ones. [6]
Automatic candidate outreach
An authorised administrator can open Recruiter profile menu → Product settings → Company Settings → Preferences → Enable Hiring Assistant auto-outreach for all dashboard users → Edit → Off → Save. This permission is documented as enabled by default; when enabled, users can configure automatic candidate messages. The default permission does not mean every project automatically sends messages. [7]
Applicant evaluation
Use Recruiter profile menu → Product settings → Preferences → Enable Hiring Assistant applicant evaluation → Edit → Off → Save. This is a separate contract-level setting, also documented as enabled by default. Project-level evaluation can additionally be paused or stopped through the Activity log. [8]
Other AI features
For writing and sales assistance, avoiding the feature prevents that particular deliberate use. It does not establish that background recommendations or processing have stopped. This review did not verify a universal Sales Navigator AI disable control or an advertising mode free of AI. Obtain product-specific confirmation where either is a business requirement.
What information and actions can the business restrict?
Decide separately what employees may publish, what they may enter into an assistant, which systems may connect, and which actions may occur automatically.
For recruitment, LinkedIn provides profile, resume-sharing and job-preference choices. It also describes an objection route for members who do not want their information processed by hiring agents, subject to applicable law. This is different from the content-model training setting. [3]
For the business acting as a recruiter, restrict the candidate information supplied to what is needed for the recruitment task. Review evaluation and outreach permissions before work begins. A summary that omits a relevant qualification can cause a suitable applicant to be overlooked; automatic contact can communicate a mistaken claim before a recruiter sees it.
Sales Navigator CRM Sync can import accounts, contacts and leads, improve recommendations, and write messages and notes back to the customer system. This integration therefore needs its own review of access and permitted changes. [5]
LinkedIn’s general Sales Navigator AI guide currently says CRM Sync data is not processed by its generative AI features. That statement does not mean CRM Sync data is unused: the integration guide describes other uses. It also does not establish the behaviour of every future connector or feature. [4][5]
A business should ask which records are synchronised, whether writing back is enabled, who can authorise it and what happens to previously imported information after disconnection. This research did not verify every connector’s revocation or deletion behaviour.
Can the business refuse use of its information for AI training?
For the covered member information, the content-model training setting provides a prospective opt-out. Optional feedback is an exception: LinkedIn says information included in feedback may still be used for training. Employees should therefore check feedback submissions as well as prompts. [2]
Regional guidance matters. LinkedIn’s later terms update says it began using some member data from the EU, EEA, Switzerland, Canada and Hong Kong for content-model training on 3 November 2025. Its UK section also describes training use. Private messages are excluded from the member training use described there. Older advice that European members are automatically outside this processing should not be relied on. [9]
Sales Navigator has separate commitments. Its AI guide says customer inputs and outputs are not used to train the models powering those features, with separate treatment for optional feedback and approval. Sessions can nevertheless be retained beyond their visible duration. [4]
The business should record which commitment applies to each route. A member account setting does not by itself establish the terms for a company Page, connected system, enterprise product or information another person supplies about the business.
How can the business check that its choices worked?
The following are proposed checks, not tests completed for this report:
- List the accounts, Pages, paid seats, agencies and integrations used for business work. Record the subscription, region, responsible person and change date.
- Confirm the training setting in each relevant member account. LinkedIn advises members who previously submitted a data-processing objection to check that their preference is reflected in the setting. [9]
- Inspect existing recruitment projects and create a harmless test project to check its initial configuration. Verify project removal separately from contract-level outreach and evaluation controls.
- Confirm that an ordinary recruiter cannot initiate the activity the business intended to prohibit. Use fictional test content where possible and avoid contacting real candidates merely to test a setting.
- Inspect connected sales systems, imported record types and write-back permissions. Compare the configuration with the business’s authorised use.
- Request retained enterprise feature information where needed. A missing session history should not be treated as evidence that the provider deleted it.
- Record the result of each check and unresolved questions. Ask LinkedIn for a written product-specific explanation where the interface does not provide enough evidence.
Settings checks demonstrate the saved preference and observable behaviour. They do not independently prove what happened inside a training process. Keep provider commitments and account verification distinguishable in the business’s records.
What does the business lose, retain or need to replace?
LinkedIn says the member training opt-out does not remove access to its generative AI features. A business can therefore refuse the covered training use while retaining feature access. [9]
Disabling recruitment assistance removes the relevant automated sourcing, evaluation or outreach. Replace it with a named recruiter, agreed selection criteria, a record of why candidates progressed and human approval of messages. Review candidates against their actual evidence rather than treating an AI summary as the complete application.
Members who limit profile information or object to hiring-agent processing may be less visible through those agents. LinkedIn says they can still apply for jobs and appear through searches that do not use the agents. This is a practical trade-off for an individual; it is not a reason for an employer to make decisions about an employee’s personal account without agreement. [3]
If the business reduces sales automation, plan time for prospect research, message preparation and record updates. Judge the replacement by its useful results: relevant prospects, accurate claims and conversations that progress, rather than simply the number of messages produced.
For public communication, a person can draft and check posts directly. The important control is approval before publication. A polished AI draft can still misstate a service, invent experience or disclose information that the business did not intend to make public.
Control assessment
| Business requirement | What this review establishes | Remaining limit |
|---|---|---|
| Refuse covered member training use | A member opt-out exists. [2] | Prospective; other model purposes and feedback need separate attention. |
| Stop recruitment automation | Project removal and contract controls exist. [6][7][8] | Check every relevant project and permission. |
| Limit automatic messages | An administrator can disable auto-outreach permission. [7] | Previously sent messages need separate handling. |
| Restrict sales-system access | CRM Sync has documented import and write-back functions. [5] | Exact permissions and disconnection effects need verification. |
| Understand retained AI sessions | Sales Navigator describes retention and enterprise requests. [4] | Closing the interface does not prove deletion. |
| Exclude member data from hiring agents | LinkedIn describes an objection route. [3] | Applicable law, outcome and visibility effects must be checked. |
| Turn off all AI across LinkedIn | No universal control established by this review. | Feature avoidance and training preferences have limited scopes. |
Official sources
[2] Control whether LinkedIn uses your data to improve generative AI models used for content creation.
[3] How LinkedIn uses AI Agents to connect job seekers and hirers.
[4] Generative AI in Sales Navigator.
[5] Sales Navigator security and privacy: CRM integration.
[6] Enable or disable Hiring Assistant in a project.
[7] Enable or disable Hiring Assistant auto-outreach for all dashboard users.
[8] Enable or disable applicant evaluation for Hiring Assistant.
[9] Update to our Terms and data use.
[10] Member training preference — sign-in required.
[11] Create an A/B test to compare Classic and Accelerate ad sets. Advertising coverage here is limited to the published description; campaign controls were not tested.
Scope and continuing research
This is a baseline of public statements and documented controls, not certification of a business account. Availability varies by subscription, assigned permissions and rollout. Some help pages could not be fetched in full; this report does not treat those pages as proof of unverified controls.
LinkedIn’s general Sales Navigator AI page contains differing language about human review and improvement in different sections. The report uses its specific customer-data commitments without claiming that all processing or review is excluded. Obtain clarification under the business’s contract if that distinction matters.
The next update should check regional training terms, Page-content treatment, advertising controls, Sales Assistant permissions, CRM integrations, retention and new recruitment defaults. LinkedIn Learning and every advertising or hiring integration are outside this report’s detailed scope.