Calculating the ROI of GEO: when the effort pays off

Why GEO ROI works differently from SEO ROI

With classic search engine optimization the chain is simple: ranking, click, session, conversion. Every step is countable, and at the end there's revenue you can trace back to a source. GEO, short for Generative Engine Optimization, breaks that chain. A user asks ChatGPT, Perplexity, or Google's AI Overview for a recommendation and gets a finished answer back. Often they never click anything. The value still gets created — your brand was named, categorized, maybe recommended.

That doesn't mean GEO has no measurable return, only that you have to change what you measure. Instead of clicks, track how often and how prominently you appear in AI answers, in what context, and with what tone. That visibility has real value because it shapes purchasing decisions long before anyone opens your website. A tax advisor, a machine shop, and a dental practice all face the same problem and the same opportunity here.

The honest answer: you won't get a cent-precise ROI the way you can with performance marketing. What you get is a robust range that tells you whether GEO is a footnote or a lever in your industry. For most businesses, that's enough to make a sound budget call.

The basic formula and its components

The formula itself stays simple: ROI equals (value gained minus costs) divided by costs, expressed as a percentage. The work is in the two numbers you plug in. Costs are the easy side — you have invoices and hours worked. Value has to be modeled, because it's made up of several indirect effects that rarely show up on a single invoice.

On the cost side this covers: time spent analyzing your visibility in AI systems, writing and revising content, technical work such as structured data, ongoing monitoring, and any tool or agency fees. Use fully loaded costs — your own time at a realistic hourly rate. Anyone who prices their own time at zero is only fooling themselves and will make bad calls.

On the value side, work your way from mention to revenue. How many relevant questions do people in your category actually ask AI assistants? In what share of those are you named? How many of those mentions turn into a visit, an inquiry, a deal? And what's a deal worth on average? Estimate each figure if you have to, as long as you write the assumption down.

Turning visibility into a monetary value

The core of any GEO calculation is turning mentions into revenue. A clean way to do it runs through what's called share of voice: the share of relevant AI answers in which you appear. Measure it by building a fixed list of typical user questions and running them regularly through several AI systems. If you're named in 20 of 100 questions, your share of voice is 20 percent.

Multiply that share by an estimated question volume and a conservative conversion assumption. Say, for a trades business: assume 500 people a month in your service area ask an AI assistant a question where a recommendation like yours could come up. At a 20 percent share of voice, you're named in around 100 of those answers. If only 5 percent turn into a real inquiry, and every other inquiry becomes a job worth 800 euros, that's roughly 2,000 euros of revenue impact a month.

Stay humble in your assumptions. Set conversion rates on the low side, because an AI recommendation doesn't replace your whole sales process. Run two scenarios — a pessimistic one and a realistic one — and make your decision based on the lower end. If GEO pays off even in the pessimistic case, the answer is clear. If only the optimistic scenario looks good, be cautious.

Naming the attribution problem honestly

The biggest weakness in any GEO calculation is attribution — tracing a result back to its cause. When a new customer calls because an AI mentioned your business, no analytics tool records that. The customer just appears to show up directly. Without a countermeasure, you'll systematically undercount GEO's contribution and credit it to other channels instead.

The simplest fix costs nothing: ask new customers how they found you, and explicitly offer AI assistant or ChatGPT as an answer on your form or in your first conversation. After a few months you'll have real data to ground your model's assumptions. You can also check your server logs for referral traffic from Perplexity, ChatGPT, or Copilot, though that only captures part of the picture.

Accept that some fuzziness will remain — that's exactly why you calculate in ranges rather than a single number. A GEO ROI of 40 to 180 percent is a more honest, more useful statement than a fabricated 87 percent. Decision-makers who understand investment risk will trust that openness far more than a false precision that falls apart under the first hard question.

A worked example

Take a mid-sized B2B software company. It invests in GEO for one quarter: 30 hours of internal time at 80 euros an hour comes to 2,400 euros, plus 3,500 euros for content and technical work, plus 600 euros in tool costs. Total: 6,500 euros. Its product needs explaining, and an average new customer brings 9,000 euros of contribution margin over the customer's lifetime. Just a couple of deals move the math significantly.

Its monitoring shows that by the end of the quarter it's named in 35 percent of relevant expert questions in AI answers, up from 8 percent at the start. It conservatively estimates monthly question volume at 300. From the added mentions, using cautious conversion assumptions, it attributes two additional deals for the quarter to AI visibility. That works out to 18,000 euros of contribution margin against 6,500 euros of cost.

That puts ROI at around 177 percent in the realistic scenario. Even in the pessimistic case, with only one attributable deal, it's still about 38 percent. Both numbers are positive, so the effort clearly pays off here. For a business with a 200-euro deal value and little AI-relevant demand, the same math could come out negative. That's exactly why the model needs to be built per industry, not borrowed from someone else's numbers.

When GEO doesn't pay off

Honesty also means naming the cases where you're better off leaving GEO alone, or investing only minimally. GEO tends to pay off poorly when your category is rarely asked about in AI assistants, when your margin per deal is very thin, or when your customers decide in a strongly local, personal way — pure walk-in traffic that does no research beforehand, for example.

There's also a timing problem. If you don't have a clean foundation yet — no structured content, no clear facts about your offering published anywhere — jumping straight to advanced GEO tactics just burns money. In that case, the first sensible investment is the foundation, not optimization. The ROI on getting the basics right is often higher than the ROI on any fine-tuning.

How to build your own calculation

You don't need an expensive tool to get started. A spreadsheet with clearly documented assumptions is enough for a first decision. What matters is that every number has a traceable origin, and that you keep optimistic and pessimistic cases separate. Update the model every quarter with real data from customer conversations and monitoring, and it gets more accurate with every round.

The process below has held up across industries. It forces you to answer the right questions in the right order, instead of getting lost in detail metrics before the basic calculation even stands up.

• Define 30 to 100 realistic user questions for your category, and measure your share of voice across several AI systems.
• Estimate monthly question volume conservatively, using search volume as a rough stand-in if you have nothing better.
• Set low conversion rates: mention to inquiry, and inquiry to deal.
• Value each deal as contribution margin over the customer's lifetime, not as one-off revenue.
• Capture every cost, including your own time at a realistic hourly rate.
• Run two scenarios, and base your decision on the pessimistic one.
• Add an origin question at first contact so you can ground attribution over time.

Separating fixed and running costs cleanly

Many ROI calculations fall apart because they lump one-time setup and ongoing operation into the same bucket. Separate the two from the start. Fixed costs cover the initial visibility analysis, building the content structure, and one-time technical work. That's a one-off sum that, for calculation purposes, should be spread across the full runtime. Charge it all against the first month and your ROI looks artificially bad early on.

Running costs are what recurs every month: maintaining content, monitoring mentions in AI answer systems, and ongoing refinement. Track these separately and honestly. A common mistake is skipping your own working time. Even if you write the content yourself instead of paying someone, every hour still costs something. Use a realistic hourly rate, or you'll end up comparing apples to oranges and overstating your return.

What period you should actually measure over

GEO doesn't work instantly. Answer systems need time to pick up new or revised content and fold it into their responses. If you calculate ROI after two weeks, you're mostly measuring noise. A meaningful measurement window usually starts only after several months, since visibility tends to build in stages before it settles at a stable level.

Set an evaluation horizon before you start, and stick to it. Three checkpoints make sense: a baseline before you begin, a check-in after roughly three months, and a full evaluation after six to twelve months. That lets you see whether the trend is rising, flat, or falling again.

Keep the same measurement approach across all three checkpoints. If you change the questions, the systems, or the counting method partway through, your numbers stop being comparable. Document your setup once, cleanly, and freeze it for the duration.

Common reasoning errors in the ROI calculation

The first common error is comparing against zero. Without GEO, you wouldn't be at zero visibility — you'd likely have picked up some mentions anyway. Subtract that baseline, or you'll credit yourself with results that would have happened regardless of the effort. Always calculate the increase over your starting point, not the absolute final number.

The second common error is inflating the revenue value per mention. It's tempting to use a high number because the result looks better. Stay conservative and use the lower end of your estimate. If the ROI is still positive under cautious assumptions, you have something solid. If it only holds up with optimistic figures, the calculation won't survive scrutiny.

The third common error is collapsing everything into a single metric. A positive ROI on its own says nothing about the variance behind it. Pair your calculation with a rough range, so you know how much your result shifts if one assumption turns out to be off.