AI agents for matched and custom audiences: the complete guide

How a list becomes an ad audience: what each platform matches on, the pipeline from rows to a live audience, retargeting, suppression and ABM audience types, the hygiene that decides match rates, and the approval gate that belongs in front of every ad account.

GuideBY THE ASTROFABRIC TEAM · SEP 1, 2026 · 11 MIN READ

A matched audience is a list you already own - accounts, contacts, customers, churned users - handed to an ad platform so it can find those same people among its members and show them, or hide from them, your ads. It is the point where go-to-market data and paid media meet, and it is where a well-built list earns a second return: the same rows that fed the sequence become the retargeting audience, the suppression list and the ABM target set. It is also where list quality shows up as a number, because the platform reports how many of your rows it matched.

This guide is the operating manual for that motion: what each platform matches on, the pipeline from rows to a live audience, the four audience types worth building, the hygiene that decides match rates, the refresh cadence that keeps audiences honest, and the approval gate that belongs in front of anything landing on an ad account. It reflects how the audience agent builds from a list.

What a matched audience is

The names differ by platform - LinkedIn calls them Matched Audiences, Meta calls them Custom Audiences from a customer list, Google calls the mechanism Customer Match - and the mechanism is the same. You upload identifiers, the platform hashes or accepts hashed values, compares them against its member records, and creates an audience containing the members it recognized. You never learn which individuals matched; you learn the size and, on some platforms, the match rate. From there the audience is targetable, excludable, or usable as a seed for the platform's own lookalike expansion. The glossary entries for matched audiences and Google Customer Match carry the short definitions; the rest of this guide is about building them well.

Platform by platform: what each one matches on

WHAT EACH PLATFORM MATCHES ON, AS OF SEPTEMBER 2026 (VERIFY MINIMUMS ON THE PLATFORM'S OWN HELP PAGES)
PlatformAudience productMatch keysNotes
LinkedInMatched Audiences: company lists and contact listsCompany name, domain, LinkedIn page URL; email, first and last name, title, companyThe only major platform that matches at the company level; audiences need a documented minimum member count to serve
MetaCustom Audiences from a customer listHashed email, phone, name, city, state, ZIP, country, date of birth, mobile advertiser IDMore identifiers per row raise the match; the list must be data you have permission to use
GoogleCustomer MatchHashed email, phone, first and last name with country and ZIP, mobile device IDPolicy-gated by account standing; audiences carry a documented minimum size before they serve on Search, YouTube and Gmail
RedditCustom Audiences from a customer listHashed email, mobile advertiser IDUseful for retargeting and suppression on community placements
XCustom Audiences: listEmail, phone, handle, mobile advertiser IDHandles are a rare identifier worth collecting when you have them
TikTokCustom Audiences: customer fileHashed email, phone, mobile advertiser IDStrong for consumer-shaped lists; B2B match rates run lower
PinterestCustomer list audiencesHashed email, mobile advertiser IDMostly retargeting and suppression for commerce lists

Two consequences fall out of the table. LinkedIn is where B2B account lists belong, because it is the platform that matches companies rather than people, so a 300-row ICP list becomes a targetable audience even when you hold no personal emails. Everywhere else, the row has to carry personal identifiers, which is why the person-data side of the list - the emails and phones found and verified upstream - decides whether an audience is large enough to serve at all.

From list to live audience

The pipeline is the same on every platform, and each stage has a way to go wrong. Normalize: lowercase and trim emails, format phones to E.164 with the country code, split names into the fields the platform expects, resolve company names to the form and domain the platform will recognize. Platforms hash what you upload; a stray capital letter or a trailing space hashes to a different value and matches nothing. Hash where the platform expects pre-hashed values, using the algorithm it specifies, so the raw data never leaves your side. Shape the file to the platform's template - column names, ordering, encodings differ and a wrong header silently drops a column. Push through the platform's API on a connected account, or export the platform-ready file where the API is unavailable to you. Wait: matching takes hours on most platforms, up to a couple of days on some. Read the match: size and rate, recorded against the list version that produced them.

The match rate is a data-quality score in disguise
A low match rate is rarely the platform's fault. It is unverified emails, single-identifier rows, personal addresses on a business list, formatting drift, or a segment the platform's membership simply does not cover. Reading the rate per list version turns the audience program into a feedback loop for the list program.

The four audience types

AUDIENCE TYPES AND WHAT FEEDS THEM
TypeBuilt fromUsed for
RetargetingContacts in sequences, open opportunities, signal movers, engaged leadsStaying visible to people already in a conversation with you
SuppressionCustomers, active opportunities, competitors, employees, unsubscribesNever paying to reach people you must not, and cleaning the lookalike seed
ABM account listThe ICP list, a signal-qualified segment, a tier of named accountsCompany-level targeting on LinkedIn; contact-level everywhere else once people are found
Lookalike seedClosed-won customers, best-fit accounts, highest-scored contactsThe platform's own expansion, only as good as the seed and its suppressions

Suppression is the type teams skip and the one that pays fastest: every impression served to a customer, an employee or a competitor is wasted money, and every one of them polluting a lookalike seed teaches the platform the wrong shape. The scoring, verification and hygiene guide covers the suppression logic that runs upstream of the audience. ABM lists inherit the executable ICP directly: one definition, consumed by the list, the sequence and the audience alike.

Match rates: the hygiene that decides them

Match rates are decided before the upload. Verified business emails match; guessed patterns and personal addresses on a B2B list mostly do not. Multiple identifiers per row - email plus phone plus name and location - give the platform more than one chance to recognize a member, and the platforms that accept several keys reward it visibly. Company-level lists on LinkedIn should carry the domain and the LinkedIn page URL as well as the name, because company names are ambiguous and domains are not. Minimum sizes must be met with room to spare, because the audience the platform builds is the matched subset, and a 1,200-row upload that matches at 60% is a 720-member audience.

Before any upload
  • Emails verified deliverable; personal addresses removed from B2B lists
  • Phones in E.164 with country codes; names split and normalized
  • Company rows carry name, domain and page URL
  • Suppressions applied: customers, opportunities, competitors, employees
  • Row count clears the platform minimum after an expected match rate
  • Consent and source documented for the platform's policy

60-80%match rates a verified, multi-identifier B2B list typically reaches on the platforms that match on personal identifiers

Refresh cadence: audiences decay

An audience is a snapshot of a list, and the list moves: people change jobs, opportunities close, customers churn, new accounts qualify. A retargeting audience uploaded once in January is, by June, showing ads to people who left and hiding from nobody who arrived. The fix is structural: the audience is a view of a persistent list, and the list is refreshed on a schedule, so the audience rebuilds from the current rows - new members added, departed members removed - without a person exporting a file. Suppression lists need the fastest cadence, because a new customer served your ads for three months is a visible mistake; ABM lists refresh with the ICP; lookalike seeds refresh when the closed-won set materially changes.

The approval gate in front of the ad account

An audience landing on an ad account is a consequential write: it changes who sees paid media and it commits data to a third party under a policy. It deserves the same treatment as a CRM push. The audience is staged with its exact membership count, its platform, its type and the list version it came from; a person sees precisely what would be created and approves or denies in one click; the approved push runs through the same governed path with the same audit entry. Once the weekly suppression refresh has been approved a dozen times, that class of push has earned autonomy, and review stays on the pushes that genuinely need it - the pattern from approval queues that keep autonomy fast.

How AstroFabric does it

The audience agent builds matched and custom audiences from any persistent list on connected LinkedIn, Meta and Reddit accounts, and produces platform-ready exports for Google, TikTok, X and Pinterest where a connection is unavailable. It normalizes and hashes the identifiers, shapes the file to each platform's template, applies the suppression lists the verification agent maintains, and stages the push for approval with the membership count, the platform and the list version shown. Retargeting, suppression and ABM audiences rebuild from the live list on the cadence you set, and the match sizes and rates land in the audit log against each list version.

Every push to an ad account passes an approval gate, every build runs under a credit ceiling, and the same list that fed the audience feeds the sequence, the CRM and the signal watch. The AI agents for audiences and agentic AI for ad targeting pages map the motion from both ends.

Go deeper in this cluster

  • What is a matched audience? - A matched audience is an ad audience built from a list you own, matched by the platform against its members. The definition, what the platforms match on, the four audience types, and why the match rate is a data-quality score in disguise.
  • What is Google Customer Match? - Customer Match is Google Ads’ mechanism for building audiences from your own customer data - hashed emails, phones, names and addresses - across Search, YouTube, Gmail and Display. What it matches on, the policy and size requirements, and how B2B teams use it well.

Frequently asked questions

What is a matched audience?

An ad audience built from a list you own: you upload identifiers such as hashed emails, phones, names or company domains, the platform matches them against its members, and the members it recognizes become a targetable or excludable audience. LinkedIn calls them Matched Audiences, Meta Custom Audiences, Google Customer Match.

Which platform is best for B2B account lists?

LinkedIn, because it matches at the company level on names, domains and page URLs, so an ICP list becomes an audience without personal identifiers. Every other major platform needs personal identifiers per row, which makes the person-data side of the list the limiting factor.

Why is my match rate low?

Usually the list: unverified or personal emails on a B2B upload, single-identifier rows, formatting drift before hashing, or a segment the platform’s membership does not cover. Verified emails, multiple identifiers per row and clean normalization move the rate more than anything on the platform side.

What is a suppression audience?

A list of people you must not pay to reach - customers, open opportunities, competitors, employees, unsubscribes - excluded from campaigns and removed from lookalike seeds. It is the audience type teams skip most often and the one that pays back fastest.

How often should audiences be refreshed?

On the cadence of the list underneath them: suppression lists weekly or faster, retargeting audiences with the sequence and opportunity data, ABM lists with the ICP, lookalike seeds when the closed-won set materially changes. Rebuild from a persistent list rather than re-uploading files.

Can an AI agent push audiences to my ad accounts?

Yes, with an approval gate in front of the ad account: the agent builds and stages the audience with its exact membership, platform and list version, a person approves in one click, and the push runs through a governed path with an audit entry. Repeated pushes of the same class can earn autonomy over time.

Sources

⟨ RUN IT INSTEAD OF READING IT ⟩

Every playbook on this blog ships as a runnable mission.

Open a workspace and the playbook library is waiting - describe the outcome and the agents carry it end to end, on your plan's monthly credits.

⟨ KEEP READING ⟩
GuideAudiences & ad targeting

What is a matched audience?

A matched audience is an ad audience built from a list you own, matched by the platform against its members. The definition, what the platforms match on, the four audience types, and why the match rate is a data-quality score in disguise.

Sep 1, 2026 · 4 min read
GuideAudiences & ad targeting

What is Google Customer Match?

Customer Match is Google Ads’ mechanism for building audiences from your own customer data - hashed emails, phones, names and addresses - across Search, YouTube, Gmail and Display. What it matches on, the policy and size requirements, and how B2B teams use it well.

Sep 1, 2026 · 4 min read
GuideAgentic GTM

The complete guide to agentic AI for GTM data

What changes when agents own the go-to-market data work: the eight data jobs, the anatomy of a data mission, the seven specialist agents, the governance that makes autonomy safe, and how to adopt it without betting the quarter.

Sep 1, 2026 · 12 min read