Forecasting ad revenue for an AI chatbot means modeling five variables — daily active users, sessions per user, ad opportunities per session, fill rate, and eCPM — because no single industry-wide dollar figure applies to every chatbot in 2026. The number most forecasts skip is fill rate: the share of ad opportunities an ad network actually fills with a paying ad, and a low fill rate can cut a rosy DAU-based projection in half before any revenue lands.
- Forecasting ad revenue for an AI chatbot runs on DAU x sessions x ad opportunities x fill rate x eCPM, not a flat number.
- Fill rate is the variable most forecasts skip - a weak fill rate can cut projected revenue in half.
- Elo's ad SDK reports fill rate and eCPM directly, so the forecast model runs on real numbers instead of guesses.
- Chatbot vertical - finance, shopping, travel, gaming - changes advertiser demand and therefore the eCPM side of the formula.
- Rebuild the forecast monthly through 2026 as fill rate and advertiser demand shift with chat volume.
Why this matters
Most chatbot builders forecast ad revenue the way mobile app developers forecast interstitial revenue: multiply users by an assumed CPM and call it done. That model breaks in conversational AI because there's no fixed ad slot — an ad only shows when the Elo matcher finds a contextually relevant moment in the conversation, and that match rate is not the same as a banner refresh rate.
A forecast built on the wrong assumptions either overstates revenue to investors or undersells the monetization case to a product team deciding whether ads are worth the UX tradeoff. Getting the formula right matters more than getting a single number right, because the formula tells you which lever to pull when the number disappoints.
How do you forecast ad revenue for an AI chatbot?
Ad revenue for a chatbot forecasts as:
Monthly revenue = DAU x sessions per user x ad opportunities per session x fill rate x eCPM / 1,000 x days in month
Each variable comes from a different place — some from your own product analytics, some from the ad network's reporting dashboard.
| Variable | What it measures | Where the number comes from |
|---|---|---|
| DAU | Unique users chatting per day | Product analytics |
| Sessions per user | Chat sessions per user per day | Product analytics |
| Ad opportunities per session | How many times the matcher can surface an ad in one session | SDK ad-load config |
| Fill rate | Share of ad opportunities actually filled with a paying ad | Ad network reporting |
| eCPM | Revenue per 1,000 filled ad impressions | Ad network reporting |
Run the formula with your own DAU and a conservative fill rate assumption first, then swap in real fill rate and eCPM data once ads are live. Measuring ad revenue per user is the fastest way to sanity-check whether your assumptions match what the network actually delivers once traffic starts flowing.
Pricing models change which side of the formula moves
Ad networks price chatbot inventory three ways, and each one shifts a different part of the formula:
- CPM (cost per thousand impressions) — revenue depends on impression volume, so DAU and sessions per user matter most.
- CPC (cost per click) — revenue depends on click-through rate on top of impressions, so ad relevance matters more than raw volume.
- CPA (cost per action) — revenue depends on a completed conversion downstream, so audience quality outweighs traffic size.
- Revenue share — a negotiated cut of advertiser spend, common in direct deals rather than programmatic fill.
Most programmatic ad SDKs, including Elo's, blend CPM, CPC, and CPA demand in the same auction so the eCPM in your formula already reflects whichever pricing model won that impression. That's why pricing ad inventory in a conversational AI app correctly means setting a price floor rather than picking one pricing model and locking it in.
Why ad revenue forecasts vary
Two chatbots with identical DAU can land on very different revenue numbers. The gap almost always traces back to one of these:
- Chatbot vertical — a finance or shopping assistant sits closer to high-intent commercial queries than a general-purpose companion app, which pulls in different advertiser demand.
- Ad load — how many ad opportunities the matcher is allowed to surface per session; more opportunities raise revenue but risk chat UX if tuned too aggressively.
- Context match quality — a matcher that surfaces an ad against the actual topic of the conversation fills more of those opportunities than one guessing off generic triggers.
- Session depth — longer, more turn-heavy conversations create more natural ad moments than short one-shot queries.
- Geographic mix — advertiser demand and spend levels differ by region, which moves eCPM independent of anything the chatbot does.
- Time of year — advertiser budgets move seasonally, so a forecast built on a single month of data understates or overstates the annual number.
A chatbot revenue forecast without a fill rate assumption is a spreadsheet wearing a guess.
“A chatbot revenue forecast without a fill rate assumption is a spreadsheet wearing a guess.”
Model your chatbot's ad revenue
See real fill rate and eCPM data instead of forecasting blind.
How much does an AI chatbot make per user?
Revenue per user depends entirely on the same formula — sessions per user, ad opportunities per session, fill rate, and eCPM — so there's no fixed per-user figure that applies across chatbots. The closest thing to a standard metric is ad revenue per daily active user (ARPDAU), tracked the same way mobile apps track it, and it only becomes meaningful once you have live fill rate data instead of assumptions.
What's a good fill rate for chatbot ads?
There's no universal "good" fill rate because it depends on ad load, matcher tuning, and how many advertisers are bidding into a given vertical in 2026. A fill rate that looks low in a niche vertical can still produce solid revenue if eCPM in that vertical runs high, which is why fill rate and eCPM need to be read together, never in isolation.
Does the LLM provider (OpenAI, Anthropic, custom) change the forecast?
The underlying LLM provider doesn't change the ad revenue formula — DAU, sessions, ad opportunities, fill rate, and eCPM all work the same whether the chatbot runs on OpenAI, Anthropic, or a custom model. What changes is integration effort: an SDK built to work across providers, like Elo's, lets a forecast model stay valid even if the underlying model changes later.
FAQ
How do you forecast ad revenue for an AI chatbot?
Ad revenue for an AI chatbot forecasts by multiplying daily active users, sessions per user, ad opportunities per session, fill rate, and eCPM, then dividing by 1,000. Run the formula with conservative fill rate assumptions before launch, then replace assumptions with real ad network data once ads go live.
What inputs does an ad revenue forecast need?
An ad revenue forecast needs DAU, sessions per user, ad opportunities per session, fill rate, and eCPM. The first three come from product analytics; fill rate and eCPM come from the ad network's reporting dashboard once ads are running.
Why do two chatbots with the same DAU earn different ad revenue?
Two chatbots with the same DAU earn different revenue because of differences in vertical, ad load, context match quality, session depth, and geographic mix. A finance or shopping assistant typically sees stronger advertiser demand than a general-purpose companion app.
Is CPM or CPC better for forecasting chatbot ad revenue?
Neither CPM nor CPC alone is better for forecasting - most programmatic ad SDKs blend both pricing models in the same auction, and the eCPM in your formula already reflects whichever model won each impression. Forecast off blended eCPM rather than picking one pricing model in isolation.
Does ad load affect the accuracy of a revenue forecast?
Ad load directly affects forecast accuracy because it sets the ad opportunities per session variable in the revenue formula. Raising ad load too aggressively can inflate the forecast on paper while degrading chat UX enough to reduce sessions per user.
How often should you update an ad revenue forecast?
Update an ad revenue forecast monthly through 2026, since fill rate and advertiser demand shift as chat volume and seasonality change. A forecast built on a single month of early data will understate or overstate the annual number.
Can you forecast ad revenue before launching ads?
Yes - forecast ad revenue before launch using your existing DAU and sessions-per-user data paired with a conservative fill rate and eCPM assumption. Replace those assumptions with real ad network numbers as soon as ads go live to correct the model.
Does the chatbot's LLM provider change the ad revenue forecast?
No - the LLM provider doesn't change the ad revenue formula itself, since DAU, sessions, ad opportunities, fill rate, and eCPM apply the same way regardless of provider. What matters for the forecast is ad load and context match quality, not whether the model is from OpenAI, Anthropic, or a custom build.
One last thing
The single biggest forecasting mistake isn't picking the wrong eCPM assumption — it's forecasting off DAU alone and never rebuilding the model once fill rate data exists. A chatbot's own fill rate, measured against its own vertical and ad load, replaces every industry-average guess a forecast started with, and that's the point where the forecast stops being a guess and starts being a plan.
