LinkedIn Ads for B2B: how the platform actually works.

LinkedIn Ads are sold in an auction where your bid and your ad's relevance decide who wins, and you target people by who they are: job title, company, seniority, skills. Most objectives charge by impression unless you choose manual bidding. An ad set needs at least 300 members to run, LinkedIn suggests 50,000 or more, and conversions are counted on a 90-day click, 90-day view window by default.

That last number surprises people, and it is one of several things that changed in the last two years. LinkedIn renamed its campaign structure, retired lookalike audiences and target cost bidding, and added AI audience types, and Audience Expansion is now on by default. A lot of advice written before 2025 is now partly wrong. This guide is the platform as it works today, checked against LinkedIn's own help center, with our read on what it means for a B2B account.

The structure: campaign, ad set, ad

Starting in October 2025, LinkedIn renamed two of its four levels. What used to be a campaign group is now a campaign, and what used to be a campaign is now an ad set. The ad account and the ad keep their names. The URL tracking parameters were renamed to match, so check any UTM templates that use the old names.

If you read a guide that talks about "campaign settings" for targeting and bidding, it means what LinkedIn now calls the ad set. A campaign can hold up to 2,000 ad sets, or 50 when you turn on campaign-level budget optimization (Dynamic Group Budget). Two things are locked once an ad set launches: its objective and its ad format. An ad set also can't be moved to another campaign, only duplicated.

Objectives, and what you actually pay for

LinkedIn groups its objectives by funnel stage: brand awareness; consideration (website visits, engagement, video views); and conversions (lead generation, website conversions, plus two recruiting objectives). The objective decides which formats you can use, which optimization goals are available, and how you are charged.

Sponsored Messaging (message and conversation ads) is the exception: it is charged per send. Every objective also allows manual bidding charged by impression. A video view counts after two continuous seconds with at least half the video on screen.

Bidding: three strategies, and one that was removed

There are three bidding strategies today.

• Maximum delivery (automated) spends the budget to get the most results. It is always charged by impression and has no bid or cap to set. LinkedIn warns that with a large budget and a small audience it may show the same member your ad more than once, which in B2B is the common case.
• Cost cap tells LinkedIn the average cost per result you want to stay under. It is charged by impression, fluctuates most in its first 7 days, and LinkedIn recommends running it for at least 15 days. Changing the audience, bidding strategy or creative resets its learning.
• Manual bidding works with every objective and format, and can charge per click, impression, view or send. The floor is $0.01, but LinkedIn shows a suggested range and warns when a bid is too low to deliver. For leads and website conversions, manual bidding is "enhanced": LinkedIn raises or lowers your bid per auction, aiming to keep the average at or below your maximum.

Target cost bidding has been removed. Existing target cost ad sets can finish, but reactivating one switches it to cost cap. If an older playbook tells you to set a target cost, that option is gone.

Our read: Maximum delivery needs volume to learn, and most B2B audiences are small, so it tends to buy frequency rather than reach. On a tight audience, manual bidding keeps the price per click in your hands. Cost cap makes more sense once conversions are flowing and you know what a lead is worth to you, and in B2B the cheapest lead is rarely the one that turns into pipeline.

Auction: bid plus relevance

LinkedIn says two things decide which ad appears: the bid and the ad's relevance to the member. Relevance draws on member activity, page context, the ad content and its targeting, and it is scored from engagement such as clicks, comments, likes and shares. LinkedIn says the more relevant your ad, the lower the price you pay, and that higher-relevance ads can beat higher bids. Four things set what you pay: how in demand your audience is, your bidding strategy, your objective and that relevance score.

What LinkedIn does not say today is that the auction is second-price, or what exactly you pay above the next bidder. Its own words are that you only pay what is needed to win. Plan on that, not on a formula.

Audience size and how targeting combines

An ad set needs at least 300 members to run. LinkedIn suggests at least 50,000 to drive results, and at least 300,000 for Sponsored Content and Sponsored Messaging. The size estimate is not your reach: LinkedIn says the number of members you actually reach will be lower than the target audience size, and delivery metrics such as reach and frequency are modeled, not counted.

Targeting combines in two ways. Several values inside one facet (three job titles, say) are joined with OR and grow the audience. Separate facets are joined with AND by default: a member must match one value from each, so every facet you add shrinks the audience. The current audience builder lets you switch that AND to OR when you want a union instead. We cover the arithmetic in the LinkedIn audience number is not your audience.

It helps to know where each facet comes from, because some are what members typed and some are inferred:

• Job title comes from member input plus an AI classifier, and never from third-party data.
• Job function and seniority are mapped from the job title with a fixed taxonomy. LinkedIn says no AI inference is used for that mapping.
• Years of experience is calculated from the dates on the member's positions.
• Member skills include inferred skills, drawn from the profile and from the job, employer, industry or education.
• Company name targets up to 200 companies directly. Longer account lists go through a Matched Audiences company list.

Audience Expansion is on unless you turn it off

LinkedIn now switches Audience Expansion on by default wherever it is available. It adds members who resemble your targeting, and they are not included in the forecast, but they are counted in reporting. For an ABM list or a tightly defined buying committee, that silently widens the audience past the accounts you meant to reach.

Expansion is one of four AI audience types LinkedIn now offers. The others are Auto-Targeting (built from your URL, inputs and account history, with no size estimate shown), buyer groups (pick up to three product categories to reach the decision-makers who buy them) and predictive audiences. For anything account-based, turn expansion off, and test the AI types in their own ad sets so their results don't blur into your core targeting.

Matched audiences, and what replaced lookalikes

Matched audiences are where LinkedIn gets precise. You can retarget website visitors through the Insight Tag, people who engaged with image, document or video ads, Lead Gen Form openers, conversation ad clickers, Company Page visitors and event attendees. Engagement audiences look back 30 to 365 days, and video audiences can be cut at 25%, 50%, 75% or 97% viewed. Every audience needs at least 300 members to be used.

• Contact lists need at least 300 rows and must match at least 300 members, up to 300,000 records. LinkedIn matches only against verified emails and recommends uploading 10,000 or more addresses to be safe. Contact lists can't be combined with member interests and traits.
• Company lists need at least 300 rows, up to 300,000 companies, and LinkedIn recommends 1,000 or more. A company list expires if it isn't used in an ad set within 90 days.
• In the EU, EEA, UK and Switzerland, matched audiences only match members who opted in, so expect smaller matches there.

LinkedIn discontinued lookalike audiences on February 29, 2024. Predictive audiences replaced them. You build one from a single seed: a contact or company list of 300 to 300,000 rows, or a conversion, Lead Gen Form or retargeting source with more than 300 members. It takes up to five days to reach full size and then refreshes daily, and each ad account can hold up to 200. A predictive audience built from a list can include the list itself, so exclude it if you only want new people. We cover seed choice in LinkedIn lookalikes are gone.

Formats: which one does which job

• Single image, carousel and video are the feed workhorses. Carousels take 2 to 10 cards, which can't be edited after saving. Videos run from 3 seconds to 30 minutes; LinkedIn recommends 15 to 30 seconds.
• Document ads let members read a PDF or deck in the feed, up to 300 pages. For lead generation they accept PDFs only. They suit reports and playbooks your buyers would actually read.
• Thought Leader Ads promote a public post by a person, an employee or even a creator, with their permission. Documents, polls and multi-image posts don't qualify, and they run on brand awareness and engagement objectives.
• Conversation ads are inbox messages with up to five buttons, so the reader chooses a path. Message ads are the single-button version and are still available. Both are charged per send, both now carry a Not Interested button, and LinkedIn limits how often a member receives them without publishing the number. See conversation ads.
• Event ads promote LinkedIn Events, and since April 2026 events hosted off LinkedIn too.
• Text, spotlight and follower ads are desktop-only right-rail and dynamic formats. They still exist, and spotlight and follower ads can't use Maximum delivery.

In the EEA and Switzerland, Sponsored Messaging only reaches members who opted in, which LinkedIn began asking for in August 2024. Plan message reach in Europe on that smaller base.

Lead Gen Forms

A Lead Gen Form opens pre-filled inside LinkedIn, costs nothing beyond the ad spend, and attaches to image, carousel, video, event, document, message and conversation ads. It takes up to 12 fields, including up to 3 custom questions, plus up to 5 consent checkboxes and up to 20 hidden fields for tracking. A work-email setting blocks common free email domains, and you can offer a Calendly or Chili Piper booking on the thank-you step. Leads sync to CRMs such as Salesforce and HubSpot through LinkedIn's partners. LinkedIn keeps the lead data for one year, so sync it rather than relying on downloads.

In our experience pre-filled forms convert well and qualify poorly, because submitting takes one tap. LinkedIn's answer is the qualified leads optimization goal: send back which leads qualified through the Conversions API or CRM Sync, within 30 days, and it optimizes toward those. It runs on Maximum delivery only, has a two-week learning phase, and LinkedIn recommends at least 5 qualified leads in that time.

Measurement: the defaults changed

• The Insight Tag powers conversion tracking, website audiences and Website Actions (conversions built from on-site actions with no extra code). LinkedIn now also takes conversion data from Microsoft's UET tag, deduplicated, outside the EEA and Switzerland.
• The Conversions API sends online and offline conversions from your server, directly or through partners including HubSpot, Salesforce, Segment and Google Tag Manager, deduplicated against the Insight Tag.
• The default conversion window is 90-day click, 90-day view. You can choose 1, 7, 30 or 90 days; conversions sent by the API or CSV can use 180 or 365 days for events such as MQL, SQL and purchase. Older guides quoting 30-day click and 7-day view are out of date.
• Attribution is last-touch, either per ad set or to the last ad set.
• Offline conversions upload by CSV, up to 300,000 rows; rows older than 90 days are ignored.
• The Revenue Attribution Report in Business Manager connects Salesforce, Dynamics 365 or HubSpot and reports pipeline, revenue won, ROAS and win rate against your LinkedIn activity.

A 90-day view window credits LinkedIn for anyone who saw an ad and converted within three months, which in a long B2B cycle is a lot of people. That is not wrong, but it is generous. Compare LinkedIn's numbers with your CRM on the same definition before you move budget, as we cover in the paid media attribution guide.

What we check first in a LinkedIn account

• Audience Expansion: on or off, and is that deliberate?
• LinkedIn Audience Network: is it on, and was that a choice? It extends ads to third-party apps and sites, and LinkedIn notes its impressions and clicks are not MRC-accredited.
• Conversion windows: still 90/90, and does anyone know that when reading the numbers?
• Bid strategy against audience size: Maximum delivery on a 20,000-person audience usually means high frequency.
• Lead quality: is anything feeding qualified leads back, through the Conversions API, CRM Sync or offline uploads?
• Old lookalikes: frozen since February 2024 and quietly going stale.

Common questions

What is the minimum audience size for LinkedIn Ads?

An ad set needs at least 300 members to run. LinkedIn suggests at least 50,000 to drive results, and at least 300,000 for Sponsored Content and Sponsored Messaging.

What is LinkedIn's default conversion window?

90-day click and 90-day view. You can choose 1, 7, 30 or 90 days, and conversions sent through the Conversions API or CSV upload can use 180 or 365 days for events such as MQL, SQL and purchase.

Does LinkedIn still have lookalike audiences?

No. LinkedIn discontinued lookalike audiences on February 29, 2024. Predictive audiences replaced them, built from a list of 300 to 300,000 rows or a conversion, Lead Gen Form or retargeting source with more than 300 members.

How does LinkedIn charge for ads?

Maximum delivery and cost cap are charged by impression. Manual bidding can charge per click, impression, video view or send, depending on the objective. Message and conversation ads are charged per send.