How to Forecast AEO ROI Before You Spend a Dollar

Most teams justify answer engine optimization after the fact, once the citations are already showing up. That is a hard way to get budget approved. The stronger move is to forecast the return before you spend, using inputs you can defend, so the decision to fund the program rests on a model your CFO can inspect rather than on a promise. This is different from measuring ROI after launch, and different from the pitch deck you eventually present. It is the math underneath both.

This guide gives you the forecasting model itself: the formula, the input ranges that keep it honest, a full worked example, and the pressure tests that stop it from becoming a fantasy. Build this once and you can size the opportunity for any category before committing a dollar.

Why forecast instead of just measure

Post-hoc measurement answers "did it work." A forecast answers "is it worth starting." Those are different questions with different audiences. Finance funds forecasts. The distinction also protects you: a forecast built on conservative, sourced inputs sets expectations you can beat, while an over-promised number sets you up to miss.

There is a second reason. AEO returns are not linear. Early months run negative while you build the content and structure. Returns arrive later and then compound. A forecast that models the timeline stops a skeptical stakeholder from killing the program in month two, when the model already predicted a loss there. If you are heading into that budget conversation, our guide to justifying AEO budget to a skeptical CFO covers how to frame the forecast once you have built it. This article is about building it.

The core forecasting formula

The projected value of an AEO program comes down to a chain of rates applied to a pool of demand:

Projected annual value = Query volume x AI usage rate x Target citation rate x Read rate x Conversion rate x Customer value

Then: Forecast ROI = (Projected annual value minus annual program cost) / annual program cost x 100

Each term is a number you can source or defend:

The discipline is in the sourcing. Two of these (conversion rate and customer value) you already know cold from your own funnel. Two (query volume and AI usage rate) come from research and estimation. And one, the target citation rate, is the only lever your program directly moves. Keep that framing front of mind: AEO is the work of raising the citation rate, and everything else is context.

Setting honest input ranges

A forecast is only as credible as its weakest assumption. Set each input as a range, not a point, so you can model a low, base, and high case.

• Query volume. Build a target query set of the real questions buyers ask, then export monthly volume. If "best tool for X" gets 2,400 searches and you have 50 similar questions, your pool might be roughly 120,000 monthly queries. Use question-form and comparison queries, not head terms, because those are what assistants answer.
• AI usage rate. This is the share of research shifting into AI. It is climbing fast and varies by audience, so test it hard. Model it low (say 15%), base (30%), and high (50%) rather than betting on one figure.
• Target citation rate. This is your program goal. Focused programs have reported reaching 20 to 30% mention rates within 6 to 8 weeks on their priority query set. Forecast a base of 20% and a high of 30%, and keep month-one at near zero.
• Read rate. Being cited is not the same as being read and believed. Hold this deliberately low, in the 20 to 40% range, so the model does not over-count a mention as a conversion.
• Conversion and customer value. Use your actual numbers. Do not inflate them to make the forecast look better; that is the fastest way to lose a finance reviewer.

The point of ranges is the sensitivity test. When you flex AI usage rate from 15% to 50%, you see how much of your forecast rides on a number you do not fully control. If the program only pays off at the high end of every input, it is fragile. If it pays off at the base case, it is real.

A worked example

Take a B2B SaaS product with these inputs at the base case:

Running the chain: 1,440,000 x 0.30 = 432,000 AI-borne queries per year. Times 0.20 citation rate = 86,400 cited appearances. Times 0.30 read rate = 25,920 influenced readers. Times 0.02 conversion = 518 customers. Times 6,000 = roughly 3.1 million dollars in projected annual value.

If the program costs 180,000 per year, forecast ROI is (3,100,000 minus 180,000) / 180,000, which is over 1,600%. That number is large enough to distrust, which is exactly why you run the low case next. Drop AI usage to 15%, citation rate to 10%, and read rate to 20%, and projected value falls to roughly 518,000, still a positive return but a far more sober one. Present both. The honest story is "even under pessimistic assumptions this pays back, and the base case is strong," not "this returns 1,600%."

The timeline the forecast has to show

A single annual number hides the shape of the return, and the shape is where programs get cancelled. Model it month by month:

1. Months 1 to 2. Foundational and negative. You are building content and structure with little citation yet. The forecast should show a loss here so no one panics when it arrives.
2. Months 3 to 4. Early citations land on priority queries. Returns turn positive, often in the 50 to 150% range as the first pages get quoted.
3. Months 5 to 6. Coverage widens across the query set and citation share compounds.
4. Month 7 onward. A mature program, where content, structure, and third-party signals reinforce each other, delivers the bulk of the return.

Showing this curve does two things. It sets the expectation that early months look like a loss, and it makes the case that cutting the program early forfeits the compounding phase that justified it.

Pressure-test before you present

Before the forecast leaves your desk, attack it:

• Sensitivity check. Which single input, when flexed to its low end, breaks the return? If it is AI usage rate, note that you do not control it and lean on the low-case number.
• Baseline reality. Your forecast assumes you can raise the citation rate from near zero. Confirm the starting point by running an audit first. Our walkthrough on measuring AI citation share across LLMs shows how to establish that baseline so your "before" number is real, not assumed.
• Attribution honesty. The forecast projects value, but proving it later is harder because AI rarely passes a clean referrer. Flag this openly and point to how you will measure it. The workable approaches are in our guide to attributing pipeline and ROI from AI-driven discovery.
• Cost completeness. Include content creation, technical work, tooling, and any agency fees in program cost. A forecast that understates cost to flatter ROI does not survive finance review.

From forecast to funded program

Once the model holds up, the forecast becomes the spine of your business case and, later, your scorecard. The same inputs you projected become the metrics you track: citation rate against target, influenced conversions against forecast, and cost against budget. That continuity is what makes the program defensible over time.

To see how a forecast turns into a running operation with tracked citation share, how OnlyAEO works lays out the full flow, and the content that raises your citation rate is produced by the AI Feed Engine. For a real example of the citation-rate lever moving, the FastTrackr AI case study shows a brand going from invisible to consistently cited. If you want to start with the free layer before funding anything, the free llms.txt generator is a no-cost first step, and full program options are on the OnlyAEO pricing page.

Forecast first, with conservative ranges and a visible timeline, and the budget conversation stops being a leap of faith. It becomes a spreadsheet your CFO can check.