Audience inclusion is a ChatGPT Ads targeting setting that tells OpenAI to show your campaign only to a custom audience of your own customers or prospects, matched from identifiers you upload such as emails or phone numbers. Audience exclusion does the opposite by blocking delivery to a group, while inclusion restricts delivery to the people inside your list.
If you want ChatGPT ad spend to reach known buyers instead of the whole eligible free-tier audience, inclusion is the lever you set at the campaign level. Skip it and your budget scatters across every matched conversation, so high-intent prospects compete for impressions with people who will never convert.
In ChatGPT Ads Manager, audience inclusion is the campaign-level control that limits who is eligible to see your ads to the users who match a custom audience list you provide. You build that list from first-party identifiers, OpenAI matches them against its logged-in user base, and only matched accounts become eligible for delivery. The setting sits alongside geography, platform, and context hints as one of the eligibility filters you configure before a campaign runs.
An inclusion audience has to clear a size floor before it will run. OpenAI's documentation states that audience inclusion and bid adjustments require at least 25,000 matched users, and the company recommends at least 100,000. Accepted identifiers are raw or SHA-256 hashed emails and phone numbers, plus Google Advertising IDs, and invalid, duplicate, or unmatched rows do not count.
Inclusion pairs with exclusion, which suppresses delivery and carries no size minimum, and with context hints, which describe relevant conversations without acting as exact-match keywords. Because ChatGPT Ads reports performance in aggregate and never exposes individual matches, AirOps AI Ads helps you see where your ads and competitors' ads appear across the prompts you track.
Resources: See how AirOps tracks where your ChatGPT ads and competitors appear across tracked prompts
The mechanism starts with your data and ends with a served impression, with OpenAI doing the matching in between.
Upload identifiers: You add a customer or prospect list using emails, phone numbers, their SHA-256 hashed versions, or Google Advertising IDs.
Match users: OpenAI compares your identifiers against its logged-in accounts and counts only valid, unique matches.
Clear the floor: The matched audience must clear OpenAI's minimum size before it can be used for inclusion, so small lists are rejected.
Apply at campaign level: You attach the audience to a campaign as an inclusion rule, optionally layering geography, platform, and a bid multiplier.
Serve and report: OpenAI weighs conversation context and your creative, serves to eligible matched users, and returns aggregate performance.
The report tells you how the included audience performed in total, such as clicks, spend, and cost per click. It does not tell you which individuals matched or which conversations carried your ad, so you cannot reverse-engineer a person's profile from the data.
Resources: Read the AirOps guide to ChatGPT ads targeting, creative specs, and policy
ChatGPT ads bill on cost per click, so every impression served to the wrong person is money spent against a buyer who was never in your market. Audience inclusion decides whether your spend concentrates on people you already know are worth reaching or spreads thin across the entire eligible free-tier base. For teams piloting a new and still opaque channel, that decision shapes whether the test produces a readable result or a muddy one.
Spend efficiency: Restricting delivery to a matched list keeps CPC budget on known prospects, so a small pilot budget buys signal from people who can actually convert.
Measurement clarity: When the audience is a defined list, you can compare ad outcomes against a holdout of the same population, which is the only clean way to read incrementality on a channel that hides prompt-level data.
Wasted budget risk: Without inclusion, a campaign competes for impressions in every matched conversation, so a niche product can burn its budget on free-tier users who will never buy before you notice the pattern.
Demand gen leads use audience inclusion to point ChatGPT ad budget at a qualified prospect list exported from the CRM, keeping spend on accounts already in the pipeline.
Growth marketers use audience inclusion to build a lookalike-style test by including a high-value customer segment and comparing its cost per click against open targeting.
Content strategists use audience inclusion to retarget newsletter subscribers with ChatGPT ads that extend a campaign they already run by email.
Identifier matching is the process where OpenAI compares the emails, phone numbers, or device IDs you upload against its logged-in users, and only the rows it confirms as real accounts become part of your addressable audience, which is why match rate matters more than raw list size.
A minimum audience size is the user-count floor an inclusion list has to clear before OpenAI will serve against it, which pushes you to aggregate broad first-party segments instead of narrow one-off lists that would never qualify on their own.
First-party data hygiene is the discipline of keeping your uploaded identifiers accurate, deduplicated, and consented, because invalid or duplicate rows never match and silently shrink the audience you thought you had.
Focus spend on known buyers by restricting delivery to a matched first-party list.
Adjust bids per audience with OpenAI's 0.1x to 10x multiplier to pay more for high-value segments.
Extend an existing email or CRM campaign into ChatGPT, where free and Go users now see ads.
Protect test budgets by keeping a new channel's spend inside a population you can measure.
Reuse the same hashed lists you already run on other paid platforms.
Aggregate several first-party segments into one list so the audience clears the size floor without padding it with weak data.
Hash emails and phone numbers before upload when your privacy policy requires it, since OpenAI accepts SHA-256 versions and matches them the same way.
Pair every inclusion campaign with a holdout of the same list, because that is the only way to read whether the ads drove incremental action.
Hold ChatGPT ad spend out of the EEA and Switzerland for custom audiences for now, following the current regional guidance.
Layer context hints and tight ad-group themes on top of inclusion, so relevance and audience work together instead of fighting each other.
Refresh the uploaded list on a set cadence, because stale identifiers drop out of the match and quietly erode reach.
Avoid treating inclusion as a precision tool that reaches named individuals. The audience operates only in aggregate, and match rates are never shown. A list that looks large on export can match far fewer real accounts. Plan around the matched size you can verify; the raw upload count is not what runs.
AirOps: AirOps AI Ads tracks where your ads and competitors' ads appear across the ChatGPT prompts you monitor, so you can decide which audiences and prompts justify inclusion spend.
ChatGPT Ads Manager: OpenAI's self-serve platform where you upload custom audiences, set inclusion or exclusion, and apply bid multipliers.
HubSpot: A CRM and marketing platform you can use to segment and export the first-party email and phone lists that feed an inclusion audience.
Pull your lists: Export your existing customer and prospect lists from your CRM this week, no new budget needed. Confirm each row has a usable email or phone number in a consistent format.
Clean and consent: Deduplicate the file, drop invalid rows, and verify you have the consent and privacy basis to use these identifiers for advertising.
Check the size: Estimate how many records you have before matching, and combine segments if any single list looks too small to clear OpenAI's minimum.
Load into Ads Manager: Upload the list to ChatGPT Ads Manager, choose hashed or raw identifiers, and set it as an inclusion audience on a campaign.
Launch with a holdout: Start a small campaign, hold part of the audience out for comparison, and read cost per click and conversions before you scale spend, since those early numbers show whether the matched audience actually responds.
Audience inclusion restricts a ChatGPT ad campaign to only the users who match a first-party list you upload.
It works by OpenAI matching your emails, phone numbers, or device IDs against its logged-in accounts.
The main constraint is a minimum matched-audience size, which keeps small or narrow lists from running.
The main risk is paying for reach you cannot verify, since matches and individuals are never exposed to the advertiser.
The leverage is pairing a clean, well-segmented list with a holdout so you can prove incremental value on an opaque channel.
Audience inclusion and audience exclusion are mirror controls that move in opposite directions. Inclusion narrows a campaign so it can only serve to users who match the list you upload, which concentrates spend on a defined group of customers or prospects. Exclusion does the reverse: it takes a list and blocks those users from ever seeing the campaign, which is how teams suppress existing customers, employees, or recent converters. The practical difference that trips people up is the size rule. An inclusion audience has to reach OpenAI's minimum matched size before it will run, while an exclusion list carries no such floor and can be much smaller. Most mature programs use both at once, including a high-value prospect list while excluding people who already bought, so the accelerator and the brake work together.
There is no fixed schedule, so treat list freshness as an operational habit tied to how fast your customer data changes. Audience inclusion only serves to identifiers that match OpenAI's logged-in users at match time, and identifiers decay: people change jobs, swap phone numbers, and abandon old email addresses. A practical cadence is to rebuild the list every time you run a new flight, and at minimum once a quarter for always-on campaigns. If your CRM updates continuously, sync monthly so new qualified contacts enter the audience while churned ones drop out. Watch your matched size over time, because a list that quietly shrinks below the minimum will stop serving without an obvious error. Refresh more often when you are layering bid multipliers on high-value segments, since those are the audiences where a stale match costs you the most. The goal is a list that reflects who your buyers are this month instead of last year.
Match rate varies because inclusion only counts identifiers that resolve to a real, logged-in ChatGPT account, and that overlap differs sharply from list to list. A consumer list built on personal email addresses and mobile numbers tends to match well, since those are the identifiers people use to sign up. A B2B list full of corporate email addresses often matches far lower, because many of those contacts logged into ChatGPT with a personal account instead. Data quality drives the rest: duplicates, typos, outdated numbers, and formatting errors never match and silently reduce your count. Regional mix matters too, since ChatGPT ads and audience features are not available everywhere, so contacts in unsupported regions will not match. Hashing does not change the outcome when done correctly, because OpenAI matches SHA-256 values the same way as raw ones. The upshot is that two lists of identical length can produce very different addressable audiences, which is why you should check matched size instead of uploaded size.
Only partly, and the limit is worth stating plainly. Audience inclusion lets you define which pool of users is eligible to see a campaign, so in that sense you control the boundary. You do not control who inside that pool actually gets served an impression, because OpenAI's system still decides each placement based on conversation context, your creative, the landing page, and relevance. Inclusion is an eligibility filter and not a delivery guarantee: being on the list makes someone reachable, but it does not force your ad in front of them. You also cannot see or select specific individuals, since the audience operates only in aggregate and matched people are never revealed. What you can influence is the quality of the inputs: a tight, well-matched list, clear context hints, and creative that only makes sense for your intended buyer all push delivery toward the right people. Treat inclusion as strong steering over a system you do not fully drive.
There is no published industry benchmark yet, so good is best judged against your own data and the practical minimum the platform enforces. The one hard line is that your matched audience must clear OpenAI's minimum size, so the first test of a good list is simply that it qualifies to run at all. Beyond that, a healthy consumer list often matches a large share of its rows, while a corporate B2B list matching a much smaller share can still be perfectly usable if the absolute matched count stays above the floor with room to spare. Judge quality by three things: whether the matched size clears the minimum comfortably, whether match rate holds steady when you refresh, and whether the audience actually performs against a holdout. A list that barely qualifies is fragile, because normal decay can push it below the threshold. Aim for matched headroom, steady match rates across refreshes, and measurable lift over your control group.