The conversational advertising glossary: 50 terms every AI publisher should know

Conversational advertising borrows half its vocabulary from twenty years of web ad tech and invents the other half on the fly. If you build, buy, or are simply curious about how AI chat apps make money, this is the shared dictionary. Each term below is defined in plain English and reflects how the words are used in 2026 — both on traditional web pages and on AI-native surfaces where the “page” is a streaming answer.

Ad-tech fundamentals

• Programmatic advertising — Buying and selling ad inventory through automated systems rather than human-negotiated insertion orders. Most digital ads today are transacted programmatically, in milliseconds, per impression.
• Real-time bidding (RTB) — A programmatic method where each individual impression is auctioned in real time as it becomes available, with buyers bidding before the ad renders. RTB is the engine under most open-market display and video advertising. See what real-time bidding actually is, and what changes when the surface is a streaming answer.
• OpenRTB — The IAB Tech Lab’s open standard that defines the data format for RTB bid requests and responses, so any compliant buyer and seller can transact. It’s maintained on GitHub and updated regularly with new fields and signals.
• DSP (demand-side platform) — Software advertisers and agencies use to buy impressions programmatically across many sources from a single interface, applying budgets, targeting, and bidding strategy.
• SSP (supply-side platform) — Software publishers use to offer their inventory to many buyers at once and maximize what each impression earns.
• Ad exchange — The marketplace that connects DSPs and SSPs and runs the auction matching buyers to inventory. Conceptually it’s the stock exchange of ad impressions.
• Ad server (adserver) — The system that decides which ad to show for a given request, delivers the creative, and logs the resulting impressions, clicks, and conversions. Elo is an adserver built for AI chat applications.
• Header bidding — A technique where a publisher solicits bids from multiple demand sources simultaneously before calling its primary ad server, instead of asking them one at a time. It raises competition — and yield — for each impression.
• Waterfall — The older alternative to header bidding: demand sources are offered inventory sequentially in a ranked list until one fills it. Simpler, but it leaves money on the table because lower-ranked buyers never see what higher ones declined.
• Fill rate — The share of ad requests that actually return a paid ad, expressed as a percentage. A request that gets no bid goes unfilled.
• Supply path optimization (SPO) — Buyers’ practice of deliberately choosing the most direct, transparent, efficient routes to inventory, cutting out redundant intermediaries that add cost without value.
• Revenue share — The split of ad revenue between the publisher and the platform or intermediary that helped monetize it, usually stated as a percentage to each side.

Pricing & measurement

• Impression — One instance of an ad being served and rendered to a user. It is the atomic unit most pricing and measurement is built on.
• CPM (cost per mille) — The price an advertiser pays per thousand impressions. The benchmark unit for brand and reach-oriented buying.
• CPC (cost per click) — Pricing where the advertiser pays only when a user clicks the ad, shifting some performance risk onto the publisher.
• CPA (cost per action) — Pricing where the advertiser pays only when a defined action occurs — a purchase, signup, or install. The most performance-aligned of the three, and the most demanding to deliver.
• eCPM (effective CPM) — A normalized earnings metric: total revenue divided by impressions, times one thousand. It lets a publisher compare CPC, CPA, and CPM deals on the same revenue-per-thousand scale.
• Click-through rate (CTR) — Clicks divided by impressions. A basic gauge of how compelling or relevant an ad is to the audience that saw it.
• Conversion — The completion of the advertiser’s goal action after an ad interaction — typically a sale, lead, or signup.
• Conversion rate — Conversions divided by the clicks or impressions that drove them; the headline measure of how well advertising turns attention into outcomes.
• Attribution — The method of assigning credit for a conversion to the ad touchpoints that influenced it, so spend can be judged by results rather than guesswork.
• First-price auction — An auction in which the winning bidder pays exactly what it bid. Now the dominant model in programmatic display, it puts the burden of bid optimization on buyers.
• Second-price auction — An auction in which the winner pays just above the second-highest bid. Once the norm, it has largely given way to first-price in open markets.

Targeting & privacy

• Contextual targeting — Matching ads to the content of the surrounding page or conversation rather than to a profile of the individual. It needs no personal data and has surged as third-party identifiers fade.
• Audience/behavioral targeting — Matching ads to a user based on inferred interests, demographics, or past behavior, typically assembled into audience segments.
• Intent signal — Evidence of what a user is trying to do right now — derived in chat from the actual question asked. Intent is the highest-value targeting input in conversational ads because it reflects present need, not a stale profile. See why intent beats audience.
• Cookie — A small file a website stores in the browser to remember state. Third-party cookies, used for cross-site tracking, are the ones being deprecated; first-party cookies set by the site you’re on remain common.
• Cookieless — Targeting and measurement approaches that don’t rely on third-party cookies, leaning instead on contextual signals, first-party data, and privacy-preserving techniques.
• First-party data — Information a publisher or advertiser collects directly from its own users and relationships. It has become the most durable and trusted data asset as third-party tracking declines.
• Consent — A user’s affirmative permission to process their data for defined purposes, required under regulations like the GDPR and CPRA. Handling consent and identity correctly is easier when it’s enforced server-side. See why consent belongs on the server.

AI-native & conversational

• Native advertising — Ads that match the form and function of the surrounding content so they fit the experience instead of interrupting it. Conversational ads are native by design. See what native advertising means.
• Conversational ad — An ad designed for a dialogue interface, presented as part of the back-and-forth rather than bolted onto a page. It reads as a relevant answer, not an interruption.
• In-thread ad — A conversational ad placed directly inside the message thread, in the natural flow of the exchange between the user and the AI. Format is the point: it lives where attention already is.
• Sponsored suggestion — A paid recommendation surfaced as a helpful next step or option — a product, tool, or follow-up — clearly marked as sponsored.
• Inference-time bidding — Running the ad decision while the model is forming its answer, before tokens stream to the user, rather than after a page paints. It’s the conversational analog of RTB and is what makes a native in-thread placement possible.
• Answer engine — A system that responds to a query with a direct synthesized answer instead of a list of links. AI chat apps and AI search experiences are answer engines.
• Streaming response — An AI reply delivered token by token as it’s generated, rather than all at once. Because the answer arrives incrementally, ad placement decisions have to happen before streaming begins.
• LLM (large language model) — The neural network trained on vast text that powers generative AI chat by predicting likely continuations of a prompt.
• Prompt — The input — user message plus any system instructions and context — given to an LLM to produce a response. In conversational ads, the prompt is a rich source of intent.
• Token — The unit of text an LLM reads and generates, roughly a word-piece. Models process and stream output in tokens, and usage is metered in them.
• GEO (generative engine optimization) — The practice of structuring content so it’s likely to be surfaced and cited by generative AI systems, the AI-era counterpart to traditional SEO.
• AEO (answer engine optimization) — Optimizing content to be selected as the direct answer an answer engine returns. Closely related to GEO; the emphasis is on being the chosen response, not just a citation.

Transparency & safety

• Brand safety — Keeping ads away from content that could harm a brand — violence, hate, or other clearly unsuitable material. It’s the floor: avoiding obvious harm.
• Brand suitability — A finer-grained judgment than brand safety: whether a context fits a specific brand’s values and campaign, even when the content is broadly “safe.”
• Disclosure/labeling — Clearly marking paid placements as advertising so users can tell sponsored content from organic. In conversational formats, honest labeling is essential to trust and is often legally required.
• Viewability — A measure of whether an ad actually had the chance to be seen — typically defined by how much of it appeared on screen and for how long.
• Ad fraud — Illegitimate activity that fakes ad value, such as bot traffic, fake clicks, or spoofed inventory, draining advertiser budgets.
• ads.txt — A public file publishers host to declare which companies are authorized to sell their inventory, letting buyers detect spoofed or unauthorized supply. An IAB Tech Lab supply-chain transparency standard.
• sellers.json — The complement to ads.txt: a public file SSPs and exchanges host listing the sellers and intermediaries in their supply, so buyers can trace who is involved in each impression.
• Ad Library — A public, browsable record of the ads running on a platform, supporting transparency and review. Elo publishes one at elo.ad/library.

Putting it together

These terms aren’t trivia — they’re the vocabulary you’ll use to decide how an AI chat app should make money and how to do it without breaking trust. For the full walkthrough, read the pillar guide: how to monetize your AI chat app.