Agency Briefing Standards for Conversational AI Ad Campaigns

A brief template built for keywords and audience segments fails before the first dollar is spent. Search and social briefs are organized around three fixed inputs: a keyword list or match-type structure, a defined audience segment built from demographic, behavioral, or lookalike data, and a creative spec for a fixed-format unit. None of those three fit how a conversational AI placement actually gets decided. Placement in this environment is set by a relevance-weighted, second-price auction that reads the context hint, the ad copy, and the landing page against the live conversation. A precise advertiser working with a modest budget can beat a vague advertiser with a large one. Search and social don't have that property, because spend and targeting precision tend to move together there. If a brief hands the system keywords instead of conversation descriptions, or demographic parameters instead of intent-stage definitions, the targeting engine lacks the inputs it needs to match anything correctly. The failure sits upstream of creative, upstream of budget, and upstream of every optimization lever a media team might reach for once the campaign is live.

How conversational targeting works, and what that demands from a brief

Conversational targeting matches ads to the live meaning and trajectory of a dialogue rather than to a keyword or a stored profile, so the brief has to supply inputs that describe conversational situations. A keyword records what someone typed. A conversation carries what someone is trying to accomplish, where they stand in a decision, what constraints they've already mentioned, and what they've ruled out. Microsoft Copilot's targeting model evaluates the entire session context rather than the most recent query alone, so the arc of the conversation determines relevance, not a single trigger phrase. Intent moves through phases inside a single session: a conversation that opens on a general topic can progress from informational to comparative to transactional, and the ad format and message that fit one phase won't fit the next. A brief that treats the placement as one static moment misses that movement. The ChatGPT platform offers three campaign objectives: CPM for awareness, CPC for traffic, and oCPC or oCPM, conversion-optimized and billed per click or per impression, for downstream conversion. Choosing among them depends on which intent phase the campaign is actually targeting, and that choice belongs in the brief, not left as a platform default applied after the fact.

The tier and access constraints that must be declared before any campaign begins

Before a brief can define targeting, it has to declare which users the campaign can actually reach, because the population eligible to see a conversational AI ad is narrower than most teams assume, with real consequences for strategy. On ChatGPT, ads reach only logged-in adults on the Free and Go tiers. Plus, Pro, Business, Enterprise, and Education users see no ads at all: every paid tier is ad-free. That tier structure hits B2B campaigns hard, because buyers whose employers pay for their AI tooling sit on plans where a paid ad never appears. So on that surface, organic AI citation becomes the primary way B2B demand gets captured, not paid placement. Google AI Overviews and Microsoft Copilot work differently still: eligible campaigns serve automatically, with no opt-in or opt-out control, so a team may already be advertising inside AI answers through existing search, Shopping, or Performance Max campaigns without ever having planned for it. A brief-stage audit of live campaigns is a prerequisite, not a nice-to-have. On Google's platforms, only text or Shopping ads from existing Search, Shopping, and Performance Max campaigns are eligible to show within AI Overviews, while text, shopping, local, and app ads from Search, Shopping, Performance Max, and App campaigns can also appear above or below the Overview. Not every AI Overview query triggers an ad: the query has to show buying intent, quality ads have to be available, the ad has to be relevant to the query, and the ad has to be relevant to the content the AI itself generated. These are not settings to tune once the campaign goes live. They settle whether the channel fits the audience a client is trying to reach, a decision the briefing stage has to make before any budget gets committed.

Intent-Stage Mapping in a Conversational Brief

The audience-segment section of a legacy brief gets replaced, in a conversational AI brief, by intent-stage mapping: a structured account of where in a decision process the target conversation is taking place. The conversation's phase decides both which ad fits and what it should say, so the stage itself is what gets targeted. Intent moves in sequence across a session. Informational intent means the user is researching and suits awareness content. If the user is weighing options, you have comparative intent, and it suits competitive positioning. If the user is ready to act, you have transactional intent, and it suits direct response and product placement. A keyword like "CRM for freelancers" reveals some intent but very little context. A conversation with an AI assistant built around that same need might reveal that the user is a solo consultant serving EU clients, worried about invoicing and compliance, comparing low-cost tools, and each of those details is a targeting dimension a segment label has no way to encode. The practical brief input, then, is a written scenario: what problem the user has named, what constraints they've mentioned, what they're comparing, and what decision they appear to be approaching. Mapping those three intent stages also maps naturally onto the three campaign objectives the platform offers: informational pairs with CPM and awareness, comparative pairs with CPC and traffic, transactional pairs with oCPC and conversion. The intent-stage map isn't just a targeting document. It doubles as the rationale for which objective the campaign should actually select.

Writing context hints: the craft the brief must specify and the team must be trained on

The context hint is the single highest-leverage input you have in a conversational AI campaign, and writing one well is a distinct craft a legacy brief template never asks for and never builds. Leaving it unspecified produces poor targeting, consistently. Teams that write hints as comma-separated topic lists get weak matches because the system is built to interpret conversational meaning, not keyword inventories, and a topic list doesn't describe a conversational situation. A well-written context hint describes a scenario: the problem the user is trying to solve, the point in the decision they've reached, and why the advertiser's offer fits that moment. It reads closer to a creative brief written for a human than to a keyword match type. Context hints are written at the ad group level, so a campaign spanning multiple intent stages needs a distinct hint for each one, and those hints are themselves testable: matching quality can be inferred from relevance signals in available reporting, which makes the hint both a brief input and an ongoing optimization variable. The brief should carry a context hint draft for each ad group, a plain-language rationale explaining why that hint describes the right conversational moment, and a note on which intent stage it addresses, so whoever reviews the brief can evaluate the targeting without reconstructing the logic from scratch. Most teams haven't built this skill yet, and a brief that leaves the context hint marked "TBD for the platform team" is quietly outsourcing the most important decision in the entire campaign.

What the creative brief must say when the placement sits below a model's answer

A conversational AI ad renders as a labeled card adjacent to a model's answer, not as a banner interrupting a page or a pre-roll running before content, so the creative brief has to specify how the ad earns relevance in that context, not just how it performs on its own. A ChatGPT ad is a labeled sponsored card that sits below the assistant's answer, not inside it. The model's response is not for sale. The commercial unit sits adjacent to it, and the user takes in both together. That adjacency means the creative brief has to describe the conversational moment the ad is entering: what the model's answer has likely just addressed, what the user's immediate next question or action might be, and how the ad's message connects to that without simply restating what the AI already said. Because the conversation's topic determines which ad shows up, generic creative built for a broad awareness brief will feel out of place the moment it lands inside a specific, high-intent conversation about a narrow problem, so the brief has to set the specificity level the creative is expected to match. Sensitive topic adjacency belongs in the brief, too, not left as a platform-level afterthought. Health, financial advice, and legal guidance conversations carry higher brand suitability exposure, and creative cleared for a general campaign can be inappropriate once served inside those contexts, so the brief should name which topic adjacencies are in scope and which are excluded. The landing page deserves its own line in the brief as well. The relevance auction reads the landing page alongside the context hint and the ad copy, so a landing page that doesn't match the other two undermines the targeting even when the hint and the creative are both well-written.

The measurement section of the brief must declare an attribution approach before the campaign launches

Conversational AI advertising makes last-click attribution structurally unreliable, so if a brief doesn't name an alternative measurement approach before launch, it will produce results nobody can actually evaluate. That makes the measurement section of the brief a strategic decision. A user can ask several follow-up questions inside an AI conversation, leave the chat, and convert later through an entirely different channel. A 2026 analysis using Criteo data on a sample of US retailers found that users referred from ChatGPT ads convert at a meaningfully higher rate than users from other referral channels, but because the gap between exposure and eventual conversion spans the conversation, last-click reporting does not capture that contribution. Buyers running live campaigns have flagged a parallel problem: prompt-level reporting isn't available, so teams cannot see what users were actually asking when their ads were served. The brief needs to state what will and won't be knowable going in. Two workarounds are available and the brief should specify which one applies: holdout tests, where a control group doesn't see the AI placement and the difference measures lift, and creative-level tagging, where UTM parameters tied to specific ad groups or context hints allow post-click path analysis. Reporting on the ChatGPT platform currently covers impressions, clicks, spend, CTR, average CPC, average CPM, and conversions. Copilot offers no separate placement-specific reporting, and while Microsoft Advertising's overall account reporting includes impressions, clicks, CTR, conversions, conversion rate, CPC, and ROAS, none of it is segmented to Copilot placements specifically. Google AI Overviews and AI Mode currently offer no segmented reporting at all, so campaigns serving there can't be isolated from the rest of search performance, and that limits how precisely you can attribute the channel's contribution. The brief cannot close these gaps. It can prevent a team from running a campaign without an agreed measurement approach, and deciding that approach at the brief stage is a far better position than arguing about methodology after the results come in.

The restructured brief checklist: the inputs a conversational AI campaign requires

A brief built for conversational AI advertising replaces the legacy template's keyword list, audience segment, and creative spec with five distinct input categories, each one there because this channel works in a structurally different way. The first is a surface and eligibility audit, which names which conversational AI surfaces are in scope among ChatGPT, Google AI Overviews, AI Mode, and Microsoft Copilot, confirms which user tiers are eligible to see ads on each one, and checks whether existing search, Shopping, or Performance Max campaigns are already serving inside AI surfaces and what they're currently saying. The second is the intent-stage map that replaces the audience segment, describing the conversational situation at each stage the campaign addresses, informational, comparative, or transactional, along with the campaign objective that pairs with each one: CPM, CPC, or oCPC. The third is the set of context hints that replace the keyword list: a drafted hint of up to 280 characters for each ad group, written as a conversational scenario rather than a topic list, with a plain-language rationale for why it describes the right moment and a note on which intent stage it serves. The fourth covers creative and landing page specification, which has to describe the conversational moment the ad is entering, including what the AI's answer has likely just addressed and what the user's next move might be, name which topic categories are excluded for brand suitability reasons, and confirm that the landing page URL stays consistent with the context hint and the ad copy, since the relevance auction reads all three together. The fifth is the measurement approach, declared before launch rather than negotiated afterward, naming the primary success metrics tied to the campaign objective, whether reach and frequency for CPM, click volume and post-click behavior for CPC, or downstream conversions for oCPC, specifying the attribution method as a holdout test, creative-level UTM tagging, or both, and acknowledging upfront what won't be reportable on each surface, including the absence of prompt-level data and the lack of segmented reporting for AI Overview campaigns. These five categories form the brief a conversational AI campaign actually needs, and if a team works through them before launch, it has already closed most of the gap that a legacy search or social template leaves wide open.