Actalike is the name Pinterest gives to a modelled expansion audience. An advertiser supplies a source list of people it already knows, and the platform returns a larger group of Pinterest users whose behaviour resembles that list. Every other major advertising platform calls the same construct a lookalike audience. Pinterest chose a different word because its model keys on what people do on the service, saving, searching, clicking through, rather than on demographic resemblance. The label lookalike appears nowhere in Ads Manager.

The purpose is prospecting. Customer lists and retargeting pools are finite and saturate quickly. Actalike converts a known list into a probabilistic population, which is how most advertisers buy reach on a platform where their own data covers a rounding error.

How an actalike audience is built

The build starts with a seed. Pinterest permits three sources: a customer list uploaded as a single-column CSV of emails or mobile advertising identifiers, a site visitor audience built from Pinterest tag events, or an engagement audience of people who interacted with Pins from a claimed domain. All three must reach 100 matched members first.

Two parameters then define the output. Actalike country scopes the model to a single market, so an advertiser running in five countries maintains five separate audiences off one seed. Actalike range sets the percentage, on a scale of 1 to 10. The figure describes the share of that country's Pinterest population the audience should occupy, not a share of the seed: 1 produces the tightest resemblance and the smallest pool, 10 roughly ten times the reach at correspondingly looser similarity. The source audience is excluded from the result, so an actalike never re-serves the people it was modelled from.

In the API the same object is a POST to the audiences endpoint on an ad account, with audience_type set to ACTALIKE and a rule carrying seed_id, country and percentage. Seeds are created first as their own objects, typed CUSTOMER_LIST or VISITOR. Pinterest's SDK documentation lists valid country values as the United States, Canada and the United Kingdom, a constraint the help centre does not restate and which may be narrower than the live product.

Nothing is targetable immediately. Audiences move through a status sequence of INITIALIZING, PROCESSING and then READY, with TOO_SMALL returned when that floor is not met. Pinterest states audiences are typically ready within 48 hours and occasionally take up to 72, the same window applied before a new seed reports a size at all.

One documentation conflict is worth noting. The developer reference accepts an Identifier for Advertisers, Apple's device identifier, among the customer list record types, while the help centre bars lists built on iOS activity, explicitly including IDFA data, under Apple's App Tracking Transparency framework. The two pages have not been reconciled.

From lookalike to actalike

Pinterest shipped the feature on June 14, 2016, alongside visitor retargeting and customer list targeting, and called it lookalike targeting at launch. Beta figures circulated at the time credited it with lifting reach by up to 30 times and click-through rate by up to 63 percent, numbers supplied by Pinterest rather than measured independently. The rename followed within months, and trade coverage from the period attributed it to positioning: Pinterest wanted its purchase-intent signals distinguished from Facebook's profile attributes.

The category itself was three years older. Facebook launched Custom Audiences in autumn 2012, letting advertisers match a hashed customer file against its user base, then released Lookalike Audiences on March 19, 2013 to expand those files algorithmically. Twitter followed in 2014, Snapchat in 2016.

Implementations have since diverged on the one number that matters. Meta's Marketing API accepts a similarity ratio between 0.01 and 0.20, double Pinterest's ceiling, plus a starting ratio that carves out the top slice for tiered segments. Google's Lookalike segments offer three fixed settings rather than a continuous scale, at 2.5, 5 and 10 percent of users in the target location. LinkedIn went the other way and retired lookalike audiences on February 29, 2024.

Where it sits in the targeting stack

Pinterest exposes actalike under the Find new customers targeting strategy, separating it from the Reconnect with users path holding site visitor, customer list and engagement audiences. It is one of six targeting controls, alongside interests, keywords, demographics, placement and Performance+ targeting, which the platform treats as layers, not alternatives.

That stack has been absorbing automation. Performance+, launched globally on October 1, 2024, bundles bidding, budgeting, targeting and creative into one workflow and accounted for roughly 30 percent of lower-funnel revenue by the first quarter of 2026. Underneath sits the Taste Graph, the behavioural model supplying actalike with its similarity signals. More than 72 percent of impressions occur on search surfaces, so the profile ordering organic results also decides who resembles whom.

Seeds increasingly arrive from outside. Pinterest's Epsilon Clean Room integration, announced on November 5, 2025, pushes segments built from 7,000-plus person-level attributes into Ads Manager within 24 to 48 hours, and whatever lands there can seed an actalike.

Why it matters to the marketing community

Scale is the plain answer. Pinterest reported $1.18 billion in second-quarter 2026 revenue and 640 million monthly active users on August 4, 2026. Against that base a retailer's customer file addresses a fraction of a percent, and the expansion model turns first-party data into a media plan.

Commercial pressure sharpens the point. European advertising growth slowed to 12 percent in the second quarter, 7 percent in constant currency, and average revenue per user there stood at $1.17 against $7.12 in the United States and Canada. That gap pushes Pinterest toward advertisers with weaker data maturity, including the self-serve cohort reached by Promote a Pin, for whom a modelled audience is the default.

The percentage also illustrates a trade-off common to every platform. Work on a New York Times clean room deal tested 1 percent and 5 percent segments off one seed, describing the first as the precision end and the second as reach bought at the cost of similarity.

Limitations and disputes

Seed quality governs everything downstream and is the constraint least visible in the interface. A model trained on newsletter signups will find more newsletter signups. Meta recommends source audiences of 1,000 to 5,000 despite a technical minimum of 100, an admission that the floor and the working threshold are different numbers. Pinterest publishes no equivalent recommendation.

Discrimination is the sharper dispute, and the precedent is not hypothetical. The United States Department of Justice sued Meta on June 21, 2022 under the Fair Housing Act, alleging that Special Ad Audience, the tool previously called Lookalike Audience, discriminated on the basis of race, sex and other protected characteristics when deciding who was eligible to see a housing advertisement. The settlement entered six days later required Meta to stop using it for housing by December 31, 2022, a withdrawal the company extended to employment and credit. A Variance Reduction System, agreed in January 2023, now adjusts delivery so the audience seeing a housing advertisement more closely matches the eligible one.

Pinterest addresses the same exposure by exclusion rather than remediation. Credit, employment and housing advertisements in certain markets cannot use actalike targeting at all. Eight sensitive categories, from health conditions and race to trade union membership and political affiliation, are closed to targeting outright, as is any proxy for them and anyone under 18.

Opacity is the third complaint. Neither the model nor the features it weighs are published, so the reason a particular user was included cannot be inspected or contested. Performance figures also come from the party selling the inventory, a pattern running from the 2016 beta numbers through to vendor case studies such as a 6:1 return-on-ad-spend claim built on lookalikes and suppression lists.

Not the same as

Lookalike audience is the generic industry term and the functional equivalent. The mechanics differ in ceiling and tiering, not in kind.

Customer list audience is the seed, not the expansion. It holds only matched known users, is targetable on its own, and an actalike built from it excludes everyone in it.

Engagement audience collects people who clicked, saved, commented on or viewed Pins from a claimed domain. It can seed an actalike, but never reaches anyone who has not already touched the advertiser's content.

Performance+ targeting is automated expansion applied during delivery, widening keywords and interests. Actalike is a static segment created in advance, and visible in reporting in a way in-flight expansion is not.

Recent developments

Pinterest notified advertisers on September 4, 2026 that rewritten Advertising Guidelines take effect on November 12. The actalike restriction on credit, employment and housing campaigns survives, but the accompanying demographic list does not: where the current text names age, gender and postcode as an exhaustive set, the replacement converts that list into an illustrative category. Parameters outside the original four now sit inside a restriction Pinterest defines rather than enumerates.

The same bundle revises the Ad Data Terms for connected television buying and campaign extension across third-party publishers, following the $465.1 million tvScientific acquisition that closed on February 17, 2026. Segments built inside Pinterest now travel to inventory Pinterest does not own, carrying a modelled audience beyond the surface whose behaviour produced it.

Elsewhere the category is loosening. Google converted Lookalike segments in Demand Gen from hard similarity thresholds into audience suggestions in March 2026, while keeping the narrow, balanced and broad tiers intact in Video campaigns as of September 1, 2026. Display and Video 360 is replacing legacy audience expansion with lookalike audiences, lifting the ceiling from 1 percent to 10. Pinterest's percentage control, unchanged since 2016, is now among the more explicit in the market.

Timeline

  • Autumn 2012: Facebook launches Custom Audiences, matching hashed customer files against its user base
  • March 19, 2013: Facebook releases Lookalike Audiences, establishing the modelled expansion category
  • 2014: Twitter adds a lookalike targeting equivalent
  • June 14, 2016: Pinterest launches visitor retargeting, customer list targeting and lookalike targeting
  • Later in 2016: Pinterest renames the expansion product actalike audiences
  • June 21, 2022: The US Department of Justice sues Meta over discriminatory housing advertisement delivery
  • June 27, 2022: The settlement is entered, requiring Special Ad Audience to be withdrawn from housing campaigns
  • January 9, 2023: Meta and the department agree compliance metrics for the Variance Reduction System
  • February 29, 2024: LinkedIn discontinues lookalike audiences
  • October 1, 2024: Pinterest launches the Performance+ suite globally
  • November 5, 2025: Pinterest opens an Epsilon Clean Room integration for externally built seed segments
  • March 2026: Google converts Demand Gen Lookalike segments from thresholds into suggestions
  • June 15, 2026: Display and Video 360 begins replacing audience expansion with lookalike audiences
  • September 1, 2026: Google Ads documents Lookalike segments for Video campaigns with the tiered reach controls intact
  • September 4, 2026: Pinterest notifies advertisers of rewritten Advertising Guidelines
  • November 12, 2026: The revised Pinterest Advertising Guidelines take effect

Summary

Who. Pinterest operates the model and the advertiser supplies the seed, uploaded through Ads Manager, pushed from a customer relationship system through an API partner, or delivered from a clean room such as Epsilon. Media buyers select the finished segment inside an ad group.

What. Actalike is Pinterest's modelled expansion audience, the platform's name for what the rest of the industry calls a lookalike. It takes a source audience of at least 100 matched members, applies a behavioural similarity model, and returns a segment sized as a chosen percentage of a single country's Pinterest population, between 1 and 10, with the source audience excluded.

When. Launched on June 14, 2016 as lookalike targeting and renamed within the same year. The percentage control has remained unchanged since, while the surrounding stack absorbed Performance+ automation from October 2024 and the restrictions around it were rewritten in guidelines effective November 12, 2026.

Where. Inside Pinterest Ads Manager under the Find new customers targeting strategy, and in the Pinterest API as an audience object typed ACTALIKE. Segments built there now extend to connected television and third-party publisher inventory acquired through tvScientific.

Why. First-party data addresses a small fraction of 640 million monthly users, and finite retargeting pools exhaust. Modelling turns a known list into a reachable population. The same mechanism that makes it useful, an unpublished model deciding who resembles whom, is why credit, employment and housing campaigns are barred from using it in certain markets, and why the equivalent tool at Meta ended up under a federal court order.