An ad group is the middle level of a ChatGPT Ads campaign, holding the bid and context hints that tell the system which conversations your ads appear in. It sits between the campaign, which sets objective and budget, and the ad, which holds your creative, carrying the targeting other platforms hand to keywords and segments.
How you split your ad groups decides most of your delivery, since each group pairs one set of hints with one bid per conversation type. Combine unrelated products or intents in one group and a single hint has to cover conversations that share nothing, so your ads surface in the wrong chats and spend leaks.
An ad group organizes a set of ads around one bid and one description of the conversations they belong in. Inside ChatGPT Ads, it is the only place you set targeting, because the platform has no keywords and no demographic segments. Instead you write context hints, natural-language descriptions of up to 280 characters that explain the situations and needs where your product fits.
Each ad group carries three things: a maximum bid, the context hints, and the ads themselves. The bid is a maximum cost per click (CPC) for Clicks campaigns or a maximum cost per thousand impressions (CPM) for Reach campaigns. You can also attach custom audiences here to raise or lower the bid for people who match an uploaded customer list. Everything in the group should point at one coherent conversation, because the hints apply to every ad it contains.
The campaign above it decides how much you spend and where; the ad group decides who sees the ads and in what conversational context. Strong landing-page content feeds that context, and AirOps helps marketing teams build the on-brand pages that give each ad group a relevant destination.
An ad group runs as the matching engine between your campaign settings and a live ChatGPT conversation. Here is the order it works in.
Inherit settings: The ad group takes its objective, budget, schedule, and country targeting from the campaign above it.
Set the bid: You enter a maximum CPC or CPM bid, and Ads Manager shows guidance on whether that bid is competitive.
Write hints: You add context hints describing the conversations, topics, and needs where the ads belong, the core targeting signal.
Match the conversation: When a user chats, the system weighs your hints, ads, and landing page against the conversation to decide eligibility.
Enter the auction: Eligible ad groups compete in a relevance-weighted, second-price auction, so your bid and your relevance both shape whether the ad shows.
The structure tells you which conversations and audiences each bid is chasing, so you can read performance group by group. It does not tell you the exact prompts that triggered each impression, since ChatGPT Ads reports far less query detail than keyword platforms.
Your ad group structure is the main control you have in ChatGPT Ads, and it decides whether your budget reaches the right conversations or drains on the wrong ones. Because the platform gives you no keyword report and almost no query data, the way you group ads is most of the optimization you get.
Targeting precision: Each ad group gets one set of context hints, so tight grouping is the only way to match a specific buyer intent instead of a vague average.
Wasted spend is the failure mode: Lump several buyer intents under one group and its lone hint pulls in off-topic chats, so you pay for impressions and clicks that never had a chance to convert.
Cleaner reporting: Because performance is reported per group, well-separated ad groups let you see which conversation themes and audiences earn clicks and which to pause.
SEO managers use ad groups to isolate each high-intent conversation theme so they can compare which topics ChatGPT surfaces their brand in.
Content strategists use ad groups to pair every landing page with the specific context hints that match the conversations that page answers.
Growth marketers use ad groups to control bids by audience, raising spend on matched customer lists and pulling back on cold traffic.
Context hints are the natural-language descriptions you write at the ad group level to tell ChatGPT which conversations, topics, and customer needs should make your ads eligible, drawn from how real buyers describe the problem they want solved in conversation.
The bid you set on each ad group feeds a relevance-weighted, second-price auction, so the number you enter is only a ceiling, and your ad and landing-page relevance to the live conversation decides how often you win a placement.
Granularity is how narrowly you split conversations across ad groups, and the practical rule is one group per audience-and-intent combination so that every context hint stays focused on a single kind of buyer and intent pairing.
Reach buyers inside ChatGPT at the moment they ask about a problem your product solves.
Target by conversation instead of keywords, matching intent that no keyword list captures.
Set a maximum CPC bid, with OpenAI suggesting a $3 to $5 starting bid per click for Clicks campaigns.
Adjust bids for matched customer audiences, spending more on your best-fit buyers.
Read performance group by group to see which conversation themes drive clicks.
Build one ad group per audience and intent, so each context hint speaks to a single kind of conversation.
Write hints in the language buyers actually use, because the system matches them against real chat wording.
Keep each ad group's landing page tightly matched to its hints, since landing-page relevance feeds the auction.
Start Clicks campaigns with a competitive maximum bid, then adjust once you see which groups win placements.
Use audience bid adjustments to push spend toward matched customer lists and away from cold traffic.
Separate your Reach and Clicks objectives into different campaigns, so each ad group inherits the right bid type.
Avoid stuffing every product and promotion into one broad ad group to save setup time. One hint then has to represent conversations with nothing in common, the auction gets a muddy signal, and your best offers compete against your weakest inside the same group.
AirOps: Build the on-brand landing pages and content that each ad group points to, so your destination reinforces the conversation your hints target.
ChatGPT Ads Manager: Create ad groups, set bids, write context hints, and read group-level performance inside OpenAI's native advertising platform.
Google Analytics 4: Track what happens after the click, connecting ad group traffic to on-site actions and conversions.
Map your conversations: List the 10 to 20 real questions and situations where buyers would want your product, pulled from sales calls and support chats. This needs no budget and can start this week.
Set your campaign: Create a campaign with one objective, a daily budget, your schedule, and the countries where ChatGPT Ads is available to you, since every ad group inherits these settings.
Split your ad groups: Build one ad group per audience-and-intent combination from your map, keeping unrelated offers apart so each hint stays focused.
Write hints and set bids: Add one focused set of context hints to each group and enter a maximum CPC or CPM bid for it, guided by the in-app bid strength indicator.
Match ads to pages: Give each group ads and a dedicated landing page that answer the exact conversation its hints describe, then launch and review group by group.
An ad group is where a ChatGPT Ads campaign turns a budget into targeting, through one bid and one set of hints.
You do your targeting here by writing natural-language hints, since ChatGPT Ads has no keywords or demographic segments.
The main constraint is limited reporting: you see group-level results without the exact prompts behind each impression.
The biggest risk is combining unrelated intents in one group, which wastes budget on off-topic conversations.
Your biggest edge is granularity: one ad group per audience and intent gives every hint a clear job.
The campaign controls the money and the market, while the ad group controls the match. At the campaign level you choose one objective, your budget, your schedule, and the countries where your ads can run. The ad group sits one level down and holds the two settings that decide who sees your ads: the bid and the context hints. That split matters because a single campaign can contain several ad groups, each pointed at a different conversation with its own bid. If you think of the campaign as the commercial decision and the ad group as the targeting decision, you build campaigns by objective and budget, then build ad groups by audience and intent. Keeping the two straight stops you from setting one bid for conversations that deserve different bids, and it keeps your reporting readable when you compare which groups earn clicks.
Run one ad group for every distinct audience-and-intent combination you are targeting, and let that map set the number. If you sell to three verticals across three funnel stages, that logic points to nine ad groups, each with its own hints and bid. The reason is mechanical: context hints live at the ad group level, so one group can only describe one coherent type of conversation well. When you add a genuinely new intent or audience, add an ad group for it. When two groups would carry nearly identical hints and bids, merge them, because extra groups with no real difference only fragment your reporting. Start lean with a handful of your highest-intent conversations so you can read results clearly, then expand as you learn which conversations convert. There is no delivery penalty for more ad groups, so size your groups to your buyer map instead of a round number.
Delivery varies between ad groups because each one competes in a relevance-weighted, second-price auction, and relevance changes group by group. Your bid is only one input. ChatGPT also weighs how well your context hints, ad copy, and landing page fit the live conversation, so a group whose hints closely match real chats can win impressions a higher-bidding but less relevant group cannot. Conversation volume differs too: some topics come up constantly in ChatGPT, while others are rare, so an ad group aimed at a niche intent will naturally serve less. Audience settings add more variation, since bid adjustments on matched customer lists raise or lower your effective bid for certain viewers. The practical read is to compare groups on relevance and conversation demand before you touch bids. If one group underdelivers, tighten its hints and check that its landing page answers the conversation before assuming the bid was too low.
Not exactly, and that surprises marketers coming from keyword platforms. You influence your ad group's placement, but you do not pick the conversations directly. Your main control is the set of context hints you write, describing the situations, topics, and needs where your product belongs. ChatGPT reads those hints alongside your ad copy and landing page, then decides in real time whether your group fits a given conversation. You get two more indirect controls: your bid, which sets how aggressively you compete, and audience targeting, which can narrow eligibility to matched customer lists. What you cannot do is attach exact-match keywords or block specific phrases the way search platforms allow. So the honest answer is that you steer the system, you do not select placements directly: write sharper hints, align the landing page, and split tightly focused ad groups, then let the system place you in the conversations those signals describe.
A good ad group setup is narrow enough that one context hint honestly describes every conversation the group targets. In practice that means a single audience-and-intent pairing, a hint written in the words real buyers use, a dedicated landing page that answers that exact conversation, and a bid matched to how much that intent is worth to you. A weak setup bundles several offers together, points traffic at a generic homepage, or leaves the hint field thin, which starves the auction of relevance signal. Judge each group on two things: whether it is earning clicks or impressions at a cost you can sustain, and whether those clicks go to a page that fits the conversation. If a group spends without converting, tighten the hint and the landing page before raising the bid. Clean separation and tight page-to-hint alignment are what good looks like here.