// glossary

The words of ChatGPT Ads.

A short definition, quotable as is, then what it changes for the advertiser. OpenAI terms are described as documented; Hiints terms are our own.

The channel

ChatGPT Ads: ChatGPT Ads are the sponsored ads shown inside ChatGPT, bought through the OpenAI Ads Manager.; Under a relevant answer, ChatGPT shows a placement labelled “Sponsored” that stays separate from the answer and plays no part in generating it. Only signed-in adult users on the Free and Go plans see it. US advertisers have had access since 9 February 2026, European ones since 31 August 2026. The logic resembles Search (an expressed intent, an ad that answers it), with one difference in how you steer: you describe conversations with context sentences, not keywords.; See also : OpenAI Ads Manager, Sponsored ad, Generative Engine Advertising (GEA)
Generative Engine Advertising (GEA): GEA is paid advertising delivered inside generative answer engines, with ChatGPT as the first concrete case.; Paying to appear next to a generated answer is what GEA means. GEO pursues a different goal: being cited inside the answer itself, without paying. On ChatGPT, GEA runs through OpenAI’s Ads Manager and its chain of campaign, ad group, context hints, ad. The advertiser gets a placement it controls and can measure, whereas GEO stays an influence with no guaranteed placement.; See also : ChatGPT Ads, GEO (Generative Engine Optimization)
GEO (Generative Engine Optimization): GEO is the practice of optimising a site to be cited in generative engines’ answers, without paid media.; GEO is about the inside of the answer: content, site structure and accessibility to answer-engine crawlers decide whether the brand gets quoted there. GEA buys the slot beside it, and an ad never influences the answer, so the two levers never merge. That is why Hiints’ Setup includes a GEA/GEO audit: an ad’s landing page must also stay readable by OpenAI’s agents.; See also : Generative Engine Advertising (GEA), Landing page
OpenAI Ads Manager: The OpenAI Ads Manager is the buying interface for ChatGPT ads: account, campaigns, ad groups, ads and reporting.; Account, campaign (objective, daily budget, countries), ad group (context hints, max bid), ad (title, body, link, image): that is the documented structure. The interface has been self-serve since May 2026, and the account stays the advertiser’s property, with the agency only operating it. Its native report holds just six daily columns: impressions, clicks, spend, CTR, average CPC, average CPM.; See also : Campaign, Ad group, Native report
OpenAI Select Partner: OpenAI Select Partner is the label of the OpenAI partner programme that Hiints belongs to for ChatGPT Ads.; Written in English, as is, never translated or abbreviated, the label names a programme and comes with an official badge supplied by OpenAI. All it says is that Hiints belongs to it: it carries no recommendation, no certification, no exclusivity and no privileged access to inventory. It is a verifiable fact, to be read as such and not as a performance claim.; See also : OpenAI Ads Manager, L’Indice Hiints
Sponsored ad: A sponsored ad is the ad unit labelled “Sponsored”, shown beneath a ChatGPT answer and kept separate from it.; The user reads the answer first and the ad only second, so the creative has to stand on its own. It combines a title, a short body, a square image and a link to a landing page. Selection weighs the conversation context, the ad group’s context hints, creative relevance and the bid, and the ad never takes part in generating the answer.; See also : Ad title, Ad body, Ad image
Free and Go plans: Free and Go are the two ChatGPT plans on which ads are shown; the higher paid plans display none.; A decision-maker on a paid plan will never see your ad. Ads are shown only to signed-in adult users on the Free and Go plans; Plus, Pro, Business, Enterprise and Edu subscribers see none. In B2B a senior decision-maker is often on a paid plan and therefore out of reach, which makes this a qualification criterion to settle before any launch.; See also : ChatGPT Ads, Country targeting
Self-serve: Self-serve is direct access to the OpenAI Ads Manager with no sales intermediary, opened on 5–6 May 2026.; Opening an account alone has been possible since May 2026. Before that, access went through a programme with spend minimums; self-serve removed the global minimum, though a floor of $25 per day per campaign remains. A reference point does not come with the access: OpenAI publishes no performance benchmarks.; See also : OpenAI Ads Manager, Campaign
OpenAI advertising policies: OpenAI’s advertising policies define which categories, claims and destinations are allowed, restricted or subject to review.; Every campaign goes through review before serving, on category, claims, image and landing page. Sensitive sectors (health, regulated finance, legal, insurance) fall under these policies: their eligibility is checked before any launch, never assumed. A rejection is fixed on the assets or on the destination. The Setup’s readiness audit starts there.; See also : Brand safety, Setup
Brand safety: Brand safety covers the rules that keep an ad from appearing in a context harmful to the brand or the user.; The advertiser receives neither the conversations nor users’ personal data, so brand safety is acted on upstream: precise context hints, a creative with no misleading claim, a consistent landing page. On OpenAI’s side, ad selection accounts for policy eligibility and brand safety, alongside relevance and bid.; See also : OpenAI advertising policies, Context hint

Targeting and structure

Campaign: The campaign is the top level in the Ads Manager: it carries the objective, daily budget, dates and countries.; As in Google Ads, the campaign is the level where spend is read and budgets are decided. It groups ad groups that share one business objective, hence the structural rule: one campaign, one clear objective, for example “Lead Gen France”. The minimum budget is $25 per day per campaign, handled as an average daily budget.; See also : Ad group, Campaign objective (Views / Clicks / Conversions), Country targeting
Ad group: The ad group is the middle level in the Ads Manager: it carries the context hints, the max bid and the ads.; The term exists in Google Ads too, where the ad group gathers keywords; here it gathers context hints. Each ad group covers one coherent intent family (for example “sales teams replacing an overpriced prospecting tool”) and its ads answer only that one. It is the account’s unit of analysis: an ad group with no impressions reveals an intent that is too narrow or poorly described.; See also : Context hint, Intent family, Campaign
Context hint: A context hint is a natural sentence, attached to the ad group, describing a conversation where the ad is useful. It is not a keyword.; Compare “how to cut the fuel budget of my company fleet” with “crm, sales, software”: the first is a good hint, the second a list. A good hint carries one idea and says who, when and what. Hints guide matching without guaranteeing it and are not serving rules, which rules out “only show in Paris”. They are the first lever of relevance, hence of cost; Hiints’ Setup writes one hundred of them.; See also : Ad group, Intent mapping, Relevance-weighted second-price auction
Campaign objective (Views / Clicks / Conversions): The campaign objective sets what the auction optimises for: views (CPM), clicks (CPC) or conversions.; Choosing the objective means deciding what the auction optimises, and therefore how the report is read. Views bills per impression (CPM) and serves visibility in qualified conversations; Clicks bills per click (CPC) for traffic, leads and conversion tests. Conversions, finally, presumes measurement is already in place through the Pixel or the Conversions API.; See also : Campaign, CPC (cost per click), CPM (cost per thousand impressions)
Relevance-weighted second-price auction: ChatGPT Ads’ auction mechanic: the winner pays based on the next bid, and the ad’s relevance weighs in the ranking.; A more relevant creative can win the auction with a lower bid: that is what weighting by relevance changes. OpenAI describes a relevance-weighted second-price auction, where ranking combines the bid with the ad’s relevance to the conversation, while the price paid depends on the competing bid. Creative and context hints thus become a cost lever, not a matter of taste.; See also : Max bid, Context hint, CPC (cost per click)
Max bid: The max bid is the ceiling, set at ad group level, that the advertiser is willing to pay per click or per thousand impressions.; The max bid sets the ceiling at ad group level: it applies to the click under the Clicks objective and to the CPM under Views, and the Ads Manager allows a custom bid. The price actually paid stays below that ceiling, because the auction is second-price. Take the following figures as starting references, not benchmarks: CPCs observed at the channel’s entry sit between $3 and $5, CPMs between $25 and $60.; See also : Relevance-weighted second-price auction, Ad group
Country targeting: Country targeting is the only documented geographic targeting in the Ads Manager: it is set at campaign level.; A campaign is set country by country among the open markets, including 31 European markets since 31 August 2026. No city or region targeting is documented, and a context hint does not stand in for this setting. A multi-market advertiser does well to open one campaign per country or language, so that spend can be read separately.; See also : Campaign, Context hint
Intent family: An intent family groups conversations that express the same need; it becomes an ad group.; “Alternative to X” and “how to solve Y” do not deserve the same ad, which is why we think in families rather than keywords. Common families are problem, comparison, alternative, purchase, setup, pricing, urgency, local, integration and trust. Each one becomes an ad group, with its own context hints, ad and landing page.; See also : Ad group, Intent mapping
Intent mapping: Intent mapping is Hiints’ method for rebuilding a sector’s buying conversations and turning them into ad groups and context hints.; Before the first euro is spent, the account has to be filled already: that is the job of intent mapping. The method collects intent sources (Google Ads search terms, SEO queries, sales calls, support tickets, reviews), groups them into families, ranks them by business proximity and policy risk, then turns each family into an ad group. The deliverable is an intent matrix giving, row by row, the conversation type, the hint, the angle and the landing page. It has nothing to do with keyword research.; See also : Intent family, Context hint, Setup

Creative

Ad title: The ad title is the first line of the sponsored unit, limited to 24 characters.; “Fleet fuel, cut” or “CRM for small teams” fit in twenty-four characters; a tagline does not. The limit forces a single statement: the product, or the outcome, for whom. The title weighs in the relevance assessed by the auction and has to answer the question raised in the conversation, without introducing the brand.; See also : Ad body, Sponsored ad, Message match
Ad body: The ad body is the description line of the sponsored unit, limited to 48 characters.; Under “CRM for small teams”, “Migration included, 14-day trial” adds what the title leaves out: a condition and a next step. The body always brings something new (a proof, a condition or a next step) and never repeats the title. Together they form one 72-character promise that the landing page has to honour above the fold.; See also : Ad title, Landing page
Ad image: The ad image is the square visual of the sponsored unit, PNG or JPG, between 640 and 1,200 pixels per side.; A simple visual that stays legible at small size beats a busy composition. The format is square (PNG or JPG) and the image goes through policy review like the rest of the ad. It is the easiest element to vary: Hiints’ Setup delivers the visuals, and the Creative Studio then produces them on an ongoing basis.; See also : Sponsored ad, Creative Studio
Landing page: The landing page is the page the ad links to; it has to continue the conversation, not just sell.; Clicks without conversions almost always point to a page that fails to keep the ad’s promise. Hence the rule: the most relevant page for the intent, never the homepage by default, meaning a comparison page for a comparison intent and a use-case page for a problem intent. The link carries UTMs, and the page must stay accessible to OpenAI’s agents, Pixel installed.; See also : Message match, UTM, OpenAI Pixel
Message match: Message match is the continuity between the conversation, the ad and the landing page’s first screen.; The test comes down to one question: do the ad title and the page’s H1 say the same thing to the same person? The full chain runs from ad group to conversion event (ad group → intent → title and body → landing page → conversion), and if one link changes the promise, the click is paid for and lost. The default rule: one ad group per intent, one page per ad group.; See also : Landing page, Ad title, Conversion event
Creative Studio: The Creative Studio is the ongoing production of ad visuals included in Hiints’ Starter and Scale plans.; Starter includes five visuals a month; Scale sets no limit. All follow the documented format: square, PNG or JPG, 640 to 1,200 pixels. With them, several angles per ad group can be tested without tying up the advertiser’s in-house design team, and creatives refreshed when CTR drops.; See also : Ad image, Setup

Measurement

OpenAI Pixel: The OpenAI Pixel is the script, documented by OpenAI, installed on the advertiser’s site to report conversions from ads.; A Pixel that does not fire is the first cause of “conversions not showing”, hence the test before launch. Placed on the site’s pages, it fires events such as page viewed, lead created, trial started or order created. It is the recommended starting point for measurement, which the Conversions API then hardens server-side.; See also : Conversions API, Conversion event, Landing page
Conversions API: The Conversions API sends conversions from the advertiser’s server to OpenAI, alongside or instead of the Pixel.; The Pixel depends on the browser; the Conversions API sends the event from the server, for example from the CRM or a server-side GTM, which makes late or off-browser conversions reliable. It is the second step of tracking: Pixel to start, Conversions API to hold over time. It is tested when the infrastructure allows it.; See also : OpenAI Pixel, Conversion event
UTM: UTMs are the parameters appended to the ad link to identify the source, campaign, ad and ad group in analytics.; Minimum convention: utm_source = chatgpt_ads, utm_medium = cpc or cpm, utm_campaign = campaign name, utm_content = angle or ad id, utm_term = ad group or intent family. The native report never reaches the conversation, so UTMs are the only way to tie an analytics session or a CRM lead back to a specific ad group.; See also : Landing page, Native report, Signal
CPC (cost per click): CPC is spend divided by clicks; on ChatGPT Ads, CPCs observed at entry sit between $3 and $5.; A high CPC is not first treated through the bid: it signals a relevance or competition issue on the intent. Average CPC appears in the native report and results from the relevance-weighted second-price auction, with a more relevant ad paying less. Alone, it says nothing about click quality; it is read alongside CPL.; See also : Relevance-weighted second-price auction, CPL (cost per lead), Native report
CPM (cost per thousand impressions): CPM is the cost of one thousand impressions; on ChatGPT Ads, observed CPMs sit between $25 and $60.; In the native report, average CPM is also the billing basis under the Views objective. It does not compare with a display CPM: a ChatGPT impression is shown beneath the answer to a specific question, in a signed-in conversation. The right comparison point is what you pay elsewhere for an impression of equal intent.; See also : Campaign objective (Views / Clicks / Conversions), Native report
CTR (click-through rate): CTR is clicks divided by impressions; it measures whether the ad is useful in the conversation where it appears.; OpenAI publishes no reference CTR, so the only sector benchmark possible comes from a base of operated accounts. CTR appears in the native report; a low CTR points to copy that adds little or a gap between the context hints and the ad, and the fix is rewriting title and body or tightening the hints.; See also : Ad title, Context hint, L’Indice Hiints
CPA (cost per acquisition): CPA is spend divided by the number of primary conversions: order, trial, sign-up, appointment.; Each business model has its primary event: order_created in e-commerce, trial_started in SaaS, appointment_scheduled for local services. CPA, spend divided by those conversions, does not appear in the native report; it is computed from what the Pixel, the Conversions API or the CRM report. The target CPA, set before launch, decides whether an ad group is stopped or scaled.; See also : Conversion event, CPL (cost per lead)
CPL (cost per lead): CPL is spend divided by the number of leads generated; it is the pivot metric for B2B campaigns on ChatGPT Ads.; In the HR SaaS case published by Hiints, effective CPL was around €250, with a demo-to-SQL conversion rate higher than Google Ads. CPL is computed outside the native report, from leads reported by the Pixel, the Conversions API or the CRM via UTMs, and only makes sense read with the qualification rate behind it: a low CPL on unqualified leads costs more than a higher CPL that converts into SQLs.; See also : CPA (cost per acquisition), SQL (Sales Qualified Lead), UTM
SQL (Sales Qualified Lead): An SQL is a lead that the sales team has qualified as a real opportunity after contact.; The HR SaaS case published by Hiints produced 12 SQLs, and that is the level at which a B2B account is judged. The SQL lives in the CRM, not in the Ads Manager: it is tied back to the ad through the UTMs kept on the lead, then sent as an event via the Conversions API when possible. Cost per opportunity, computed here, is one of the metrics in L’Indice Hiints.; See also : CPL (cost per lead), Conversions API, L’Indice Hiints
Conversion event: A conversion event is a measured action on the site or in the CRM, sent to OpenAI through the Pixel or the Conversions API.; The recommended standard events are page_viewed, lead_created, registration_completed, trial_started, subscription_created, appointment_scheduled, checkout_started and order_created. Each model picks one as primary and others as secondary. Without an event, reporting stops at the click: it is the last link in the ad group → intent → ad → landing page → conversion chain.; See also : OpenAI Pixel, Conversions API, CPA (cost per acquisition)
Native report: The Ads Manager’s native report gives six daily columns: impressions, clicks, spend, CTR, average CPC, average CPM.; It says how much, never why or compared with what. The native report gives no triggering queries, no placement data, no demographic breakdown and no view of the conversations that led to the display, and OpenAI publishes no benchmarks, by advertiser or by industry. UTMs, the CRM and a base of comparable accounts fill that gap.; See also : OpenAI Ads Manager, UTM, L’Indice Hiints

Hiints

L’Indice Hiints: L’Indice Hiints is Hiints’ base of sector benchmarks for ChatGPT Ads, built from its own accounts since January 2026.; OpenAI publishes no benchmarks; L’Indice Hiints exists to fill that gap. It aggregates, by sector, the CPCs, CPMs, CTRs and costs per opportunity observed on the accounts Hiints operates, on the platform since 1 January 2026. It lets you check whether your CPC sits within the sector norm or outside it, which the native report cannot do. It is proprietary, and no client account data is published individually.; See also : Native report, Le Relevé, CTR (click-through rate)
Le Relevé: Le Relevé is Hiints’ free report that positions a sector within L’Indice Hiints, delivered in 48 hours.; What do costs and intents look like for my sector on ChatGPT Ads? Le Relevé answers that one question, drawing on L’Indice’s references and a first mapping of intent families. Its 48-hour turnaround comes from the manual work it takes. It precedes the decision to launch and does not replace the Setup.; See also : L’Indice Hiints, Setup, Intent family
Signal: Signal is Hiints’ proprietary measurement layer: it links Pixel, Conversions API, UTMs and CRM to read beyond the native report.; The native report stops at the click and six columns; Signal extends the reading to the opportunity. It assembles the components documented by OpenAI (Pixel, Conversions API, UTMs) with the advertiser’s CRM data to read cost per lead, per SQL and per opportunity, ad group by ad group. Decisions can then rest on opportunities rather than clicks, and Signal also feeds L’Indice Hiints, in aggregated, anonymous form.; See also : Native report, Conversions API, L’Indice Hiints
Setup: The Setup is Hiints’ launch offer, free of charge and completed in under 48 hours: GEA/GEO audit, 100 context hints, visuals, tracking.; Free and with no commitment, the Setup covers, once, everything that precedes the first euro spent: the eligibility and readiness audit, intent mapping, the writing of 100 context hints, ad visuals, and tracking and reporting installation. It leads into Starter (15% of ad spend under €5,000 of monthly media) or Scale (10% above). The Ads Manager account belongs to the advertiser.; See also : Intent mapping, Context hint, Creative Studio

// shall we talk?

A term missing from your account?

Thirty minutes to review your structure, your hints and your tracking with the vocabulary that goes with it.

Book a call

30 minutes · no commitment · zero pitch