LinkedIn got rid of lookalike audiences. What's the best way to target companies like my closed-won customers?
LinkedIn retired lookalike audiences on February 29, 2024. The closest replacement is a predictive audience: upload your closed-won accounts as a company list (or their buyers as a contact list), and LinkedIn's AI finds members likely to act like them. The list needs at least 300 rows and 300 matched members. If you have fewer closed-won deals, build the profile yourself: find the industries, company sizes and job functions that close, then target those attributes or upload a list of similar accounts. Audience expansion adds reach, but it isn't a lookalike.
As of October 2026.
What happened to LinkedIn lookalike audiences?
They were discontinued on February 29, 2024. Since then you can't create or edit a lookalike audience. Existing ones stopped refreshing and became static. Ad sets that still use one keep delivering to that frozen audience, and a lookalike audience that isn't in a draft or active ad set is archived after 30 days. LinkedIn shut down the Lookalike API on the same date, which affected partners such as HubSpot.
LinkedIn now points to two features instead: predictive audiences for contact list, company list, conversion, Lead Gen Form or retargeting data, and audience expansion for Matched Audiences and attribute targeting.
Can I build a predictive audience from my closed-won customers?
Yes, if you have enough of them. Export your closed-won accounts (or the contacts on those deals) from HubSpot or Salesforce, upload them as a company or contact list, and use that list as the source. LinkedIn builds an audience of members it predicts will act like your source data. LinkedIn's developer docs describe the model as "behavioral + lookalike."
To create one in Campaign Manager:
- Click Plan, then Audiences.
- Click Create audience and select Predictive audience.
- Name it and pick the source type: Contact list, Company list, Conversion, Lead gen form or Retargeting.
- Tick the source, choose the location of your target audience, and set the size with the slider.
- Click Agree & Create.
| Rule | What LinkedIn says |
|---|---|
| Source types | Contact or company list, conversions, Lead Gen Form, retargeting. One source type per audience |
| Minimum source | 300 member accounts. A list must have 300 to 300,000 rows and show Ready or Updating |
| Refresh | Daily. After it shows Ready, it can take up to five days to reach full size |
| Audience expansion | Turned off for ad sets that use a predictive audience |
| Limits | 200 predictive audiences per ad account, inside the 1,000 audience segment cap |
| Sharing | Can't be shared across ad accounts or through Business Manager |
LinkedIn says predictive audiences suit bottom-of-funnel campaigns. A predictive audience built from a list can include the list's own companies or contacts. If your goal is new logos, LinkedIn recommends adding the seed list as an exclusion in the ad set, so you don't spend on accounts you already have.
Is audience expansion a replacement for lookalikes?
Only partly. Audience expansion shows your ads to members with attributes similar to your target audience, and LinkedIn says it "primarily uses professional demographics." It widens targeting you already built; it doesn't learn from your customers. It's on by default where available. You turn it off by clearing Enable Audience Expansion in the ad set's Audience section. It isn't available with predictive audiences, Auto-Targeting, buyer groups or dynamic ad formats, and anything you exclude stays excluded.
What if I don't have 300 closed-won accounts?
Then build the "lookalike" yourself from your deal data. This is ICP-based targeting.
- Export closed-won and closed-lost deals with each company's industry, employee count, revenue and region, plus the job titles of the buyers.
- Compare win rates by attribute, not just counts. An industry with ten wins and two losses beats one with twenty wins and sixty losses.
- Turn the winning attributes into LinkedIn targeting. Company attributes include Company Industry, Company Size, Company Revenue, Company Growth Rate and Company Category. Job attributes include Job Title, Job Function and Job Seniority.
- Or build a list of accounts that match the profile but aren't customers yet, and upload it as a company list. LinkedIn needs at least 300 rows, recommends 1,000 or more, and accepts up to 300,000 companies. A list can take up to 48 hours to process and must match at least 300 member accounts to run.
Which approach should I pick?
| Your situation | Use |
|---|---|
| 300 or more closed-won accounts, or 300 or more matched buyer contacts | A predictive audience seeded with that list |
| Fewer than 300 closed-won accounts | Firmographic and job targeting built from your win-rate analysis |
| You have a named target list | A company list upload, with audience expansion off |
| A tight audience that needs more reach | Audience expansion (not with predictive audiences) |
What changed in the API?
If you use a tool to manage audiences, two 2026 changes matter. From version 202607, fetching a predictive audience returns its id, so partners can find and update audiences created in Campaign Manager. From August 2026, the API caps each ad account at 1,000 audience segments, and predictive audiences count toward it; going over returns a 429 error with the code SEGMENT_LIMIT_EXCEEDED. The Predictive Audiences API needs Matched Audiences API access and doesn't accept Lead Gen Forms or conversions as sources.
Where Tofu fits
Tofu Ads Agent connects to HubSpot or Salesforce, reads your closed-won and closed-lost deals, pipeline, sales velocity and deal size (plus your last 200 live ads), and proposes an ICP and segments. It then builds LinkedIn targeting from that ICP by job title, seniority, function, industry, company size and named companies, and it can build matched contact or company audiences from a HubSpot or Salesforce list or report. Be clear on what this is: ICP-based targeting, not a lookalike model. Tofu doesn't score LinkedIn members against your customers the way a predictive audience does. It costs a flat $2,500 a month.
Sources
- LinkedIn lookalike audiences have been discontinued, checked October 23, 2026
- Create LinkedIn predictive audiences for ad targeting, checked October 23, 2026
- LinkedIn predictive audiences, checked October 23, 2026
- Use audience expansion in your ad set, checked October 23, 2026
- Targeting options for LinkedIn Ads, checked October 23, 2026
- Company list targeting in Campaign Manager, checked October 23, 2026
- Matched Audiences: Overview (Microsoft Learn), checked October 23, 2026
- Recent Marketing API Changes (Microsoft Learn), checked October 23, 2026
- Tofu Ads Agent (llms.txt), checked October 23, 2026
