GEO (generative engine optimization) ROI is measurable, but only part of it arrives as a click. Measure it in three layers: visits from AI assistants and the sales or leads they produce, which Google Analytics 4 now separates into its own AI Assistant channel; branded demand that starts in an AI answer but arrives through search; and deals where the buyer says an AI tool introduced them. Then take the gross profit you can attribute, subtract what the program cost, and divide by that cost.
Don’t build the case on someone else’s conversion rate. Published figures run from AI referrals converting less often than other traffic (Adobe’s US retail data, July 2025) to 23 times more often than organic search (Ahrefs’ own website, June 2025), and the most-quoted numbers come from a single site, a publisher network or an undisclosed sample. The table below shows what each study actually measured. The only conversion rate that belongs in your ROI model is your own.
For the tracking setup itself, see How to Measure AI Citations; for tool options and prices, Best AEO Tools in 2026.
What the published conversion data actually says
These are the studies most often quoted in GEO business cases, with the scope each one covers. All were checked on November 10, 2026.
| Who and when | What was measured | Finding |
|---|---|---|
| Seer Interactive, published June 3, 2025 | One client website, GA4 data from October 1, 2024 to April 30, 2025; conversion = GA4 key events | ChatGPT referrals converted at 15.9% vs 1.76% for Google organic (Perplexity 10.5%, Claude 5%, Gemini 3%) |
| Ahrefs (Patrick Stox), June 16, 2025 | Ahrefs’ own website only, its own web analytics, 30 days | AI search sent 0.5% of visitors and 12.1% of signups, which Ahrefs calls a 23x higher conversion rate than organic search |
| Microsoft Clarity, November 6, 2025 | Over 1,200 publisher and news sites using Clarity; one month of conversion data | Sign-up rate 1.66% from LLM referrals vs 0.15% from search; subscription rate 1.34% vs 0.55% |
| Semrush, page dated July 21, 2025 | Semrush has not published the sample or period behind this figure | The average AI search visitor is “4.4 times as valuable” as a traditional organic visitor, “based on conversion rate” |
| Adobe Digital Insights, August 21, 2025 | Generative AI referrals to US retail sites | In July 2025, AI traffic was 23% less likely to convert than other traffic, down from 49% less likely in January 2025 |
| Adobe Digital Insights, April 2026 quarterly AI traffic report | US retail sites, from Adobe Analytics data covering more than a trillion visits to US sites | In March 2026, AI-referred visits converted 42% more than non-AI visits, with 37% higher revenue per visit |
Three things stand out.
The direction has mostly been positive, but the size depends on the site. Within a year, Adobe’s retail series went from AI visitors converting worse to converting better. Seer’s 9x gap and Ahrefs’ 23x come from one website each; Clarity’s figures come from publishers selling sign-ups and subscriptions, not B2B pipeline. None of them tells you what your buyers will do.
Not every quality signal points the same way. In a separate June 2025 study of 81,947 sites, Ahrefs found AI visitors viewed fewer pages than search visitors (4 vs 5.2) and bounced more often (67.8% vs 63.7%). Higher conversion and lower engagement can both be true: some AI visitors arrive ready to act, others only to check one fact.
Volumes are still small. On Seer’s client site, all AI traffic combined was under 0.1% of organic sessions (just under 11,000 sessions against almost 14 million). AI search was 0.5% of Ahrefs’ visitors, and Clarity found AI referrals under 1% of traffic on its publisher sites. A high rate on a small base is still a small number of conversions, so report both.
The three layers of GEO ROI
Measuring only the first layer undercounts. Adding all three together overcounts, because the same deal can appear in each. Measure them separately and report them side by side.
Layer 1: Direct AI referrals
Visits that click through from an AI assistant and then buy, sign up or become a lead. This is the hardest number you have.
In Google Analytics 4. Since May 13, 2026, GA4 sorts these visits into an AI Assistant channel in the Default Channel Group, with no setup. When the referrer matches Google’s list of AI assistants, GA4 sets the medium to ai-assistant and the campaign to (ai-assistant). Google’s definition names “sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok” and does not publish the full list. Three practical consequences:
- Check what lands elsewhere. Google’s announcement also names Claude; still look in the Referral channel for
perplexity.aiand any other assistant domains. Default channel groups can’t be edited, so if any land in Referral, build a custom channel group that combines the AI Assistant channel with those sources, placed above Referral. - Mind the break in your history. GA4 sets the medium when the visit is recorded, so AI visits from before the channel reached your property stay under Referral. Compare periods on a custom channel group, not on the default one.
- Treat it as a floor, not the total. A visit that arrives without a referrer can’t be matched and is counted as Direct, and the channel excludes Google’s AI Overviews and AI Mode by definition.
Record AI Assistant sessions, key events, conversion rate, and revenue or lead value each month, alongside the same figures for organic search.
In Google Search Console. Clicks from AI Overviews and AI Mode are reported as ordinary Google Search traffic. Search Console’s Generative AI performance reports, announced June 3, 2026 and available to all websites since August 31, 2026, show how often your URLs appeared in generative AI features (impressions), which pages appeared, and in which countries and on which devices. They show impressions, not clicks, so use them as a visibility measure rather than an ROI input.
Layer 2: Assisted discovery through branded search
A buyer reads your name in a ChatGPT or Gemini answer, then searches for you on Google and converts. GA4 records branded organic search; the AI answer that started it leaves no trace.
To estimate it, track branded query impressions and clicks in Search Console month over month, and compare the trend with your citation tracking (below). Treat a rise as a directional signal, not proof: PR, ads, word of mouth and seasonality move branded search too.
Layer 3: Pipeline influence
Deals where the buyer says an AI tool introduced them or shaped the shortlist, whatever channel they last arrived through. There is no published benchmark for the share of deals that cite AI discovery, so you have to collect it:
- Add a free-text “How did you first hear about us?” field to forms, and an AI assistant option if you use a dropdown.
- Ask the same question on every sales qualification call and store the answer in a CRM field.
- Report the count and value of opportunities that mention an AI assistant, quarter by quarter.
Self-reported attribution is imperfect, but it is the only way to see an AI answer that persuaded a buyer without producing a click.
The leading indicator: AI visibility
Before any of the three layers moves, your brand has to appear in the answers. Track a fixed set of buyer prompts across the AI surfaces your customers use, and record three numbers: appearance rate (the share of answers that name you), citation rate (the share that link to your site) and share of AI voice (your mentions as a share of all mentions for you and the competitors you track). AI answers vary from run to run, so use the same prompts every time and re-run them on a schedule. How to Measure AI Citations covers the method; Competitor Analysis for AI Citations covers the competitive view.
Visibility is not ROI. It explains why the business layers move, and warns you early when a competitor starts taking your place.
How to calculate GEO ROI
Attributable value
Use gross profit, not revenue: a dollar of revenue at a 30% margin doesn’t pay for a dollar of marketing.
For e-commerce and other direct sales:
AI-attributed gross profit = AI-referred orders × average order value × gross margin
For lead generation, value the pipeline first, then discount it to expected profit:
AI-sourced pipeline = AI-referred qualified leads × lead-to-opportunity rate × average deal value
Expected gross profit = AI-sourced pipeline × win rate × gross margin
Pipeline and revenue are different numbers. Don’t put pipeline into an ROI formula as if it were closed revenue.
Cost
Include everything the program uses: agency or consultant fees, tracking tools, content production, developer time for technical fixes, and internal hours spent on reviews and approvals. For market prices, see How Much Does GEO Cost in 2026?; ours are on our pricing page.
The formula
GEO ROI (%) = (attributable gross profit − GEO cost) ÷ GEO cost × 100
Count only the value GEO caused, not every AI referral you would have received anyway (see the next section).
A worked example
The numbers below are illustrative, chosen to show the arithmetic. They are not a benchmark or a forecast.
A B2B software company averaged 1,000 AI Assistant sessions a month in the three months before its program and 1,600 a month in the latest three. Over the same period, AI sessions to a group of comparable pages it didn’t touch grew 20%, so without the program it would have expected about 1,200. That leaves 400 incremental sessions a month.
- 400 sessions × 3% conversion to a qualified lead = 12 leads
- 12 leads × 25% lead-to-opportunity = 3 opportunities
- 3 opportunities × $12,000 average deal = $36,000 pipeline
- $36,000 × 20% win rate × 80% gross margin = $5,760 expected gross profit a month
- Program cost: $4,000 a month
- ROI = ($5,760 − $4,000) ÷ $4,000 = 44%
That is Layer 1 alone. Branded-search lift and self-reported AI discovery go beside it in the report as supporting evidence, not on top of it. On a long sales cycle, the expected value appears months before the cash, so report pipeline and closed revenue separately.
Separating GEO’s effect from the channel’s growth
AI referrals are growing on their own, which makes it easy to credit a program with growth it didn’t cause. Previsible’s 2026 State of AI Discovery report (published July 6, 2026) analyzed 166 GA4 properties across SaaS, e-commerce, finance, legal, health, publishing and other sectors. Across them, monthly LLM-referred sessions grew 9.9x from November 2024 to May 2026, and ChatGPT sent 92.4% of the 6.77 million sessions in the dataset.
The same report shows how unstable the channel is. Monthly sessions fell 50% in a single month in late 2025, driven almost entirely by ChatGPT referrals dropping from 448,412 in October to 213,345 in November; total sessions were back at 442,609 in December. A program judged in November would have looked like a failure, and one judged in December like a success, for reasons that had nothing to do with either.
Three ways to isolate your own effect:
- Use control pages. Leave a comparable set of pages untouched and compare their AI referrals and citations with the pages you optimized over the same period. The difference between the two trends is a fairer estimate of the program’s effect than the raw change.
- Hold the prompt set fixed. Visibility gains only mean something when they’re measured on the same prompts, engines and schedule each time.
- Annotate everything. Mark site launches, migrations, the date the AI Assistant channel reached your property and major AI product changes on your charts, so a jump isn’t misread as a result.
Don’t project AI traffic forward in a straight line from a few months of data. Model a range, and update it as your own baseline grows.
What published GEO case studies show, and what they don’t
We haven’t found an independent, controlled study of the ROI of GEO programs. The public case studies are agencies reporting on their own websites, which makes them marketing for those agencies. They show what is possible, not what is typical.
Go Fish Digital (September 24, 2025, its own website). Over three months, the agency reports a 43% rise in monthly AI-driven traffic, an 83.33% rise in monthly conversions from those AI referrals, and AI traffic converting at “25X” the rate of traditional search.
The Rank Masters (September 12, 2025, its own relaunched website). ChatGPT-referred page views rose from 8 (March 12 to June 9, 2025) to 675 (June 11 to September 8, 2025), the 8,337% in its headline. The post’s own views-per-user figure (48.21) implies about 14 active users, so this is a percentage on a very small base.
Reporting GEO ROI to leadership
Monthly
- AI Assistant sessions and conversions from GA4 (plus any custom-group sources), with organic search beside them. Label it as a floor.
- Generative AI impressions from Search Console for your priority pages.
- Appearance rate, citation rate and share of AI voice on your fixed prompt set, against the prior month and your baseline.
- Branded search impressions and clicks from Search Console.
Quarterly
- Attributable gross profit and ROI from Layer 1, using incremental sessions, with the cost basis shown.
- AI-influenced pipeline from self-reported attribution, as a separate line.
- AI-referred conversion rate vs other channels, from your own data only.
- Share of AI voice trend against your main competitors. A falling share with flat effort usually means a competitor has stepped up.
Show the layers side by side and label which are measured and which are estimated.
When GEO ROI is hard to measure
- No conversion event on the site. Brand-led businesses, or those that sell offline, have little to count in Layer 1; rely on visibility, branded search and sales-call attribution.
- Long sales cycles. When deals take a year to close, report pipeline and opportunity counts, and accept that closed-revenue ROI lags the work.
- Mostly zero-click influence. If buyers mostly read AI answers and then call or visit in person, Layer 1 will stay small however well the program works; weight self-reported attribution more heavily.
FAQ
What is the ROI of GEO?
It depends on your margins, your sales cycle and what the program costs, so there is no industry number to borrow. Calculate it as attributable gross profit minus GEO cost, divided by GEO cost, using your own GA4 conversion data and only the AI referrals the program caused. Published studies show AI referrals converting anywhere from worse than other traffic (Adobe, US retail, July 2025) to far better (Ahrefs, its own website, June 2025), on small volumes.
How do I track AI traffic in Google Analytics 4?
Use the AI Assistant channel GA4 added to the Default Channel Group on May 13, 2026. GA4 assigns it automatically when the referrer matches a recognized AI assistant; Google’s definition names sources like ChatGPT, Gemini, Deepseek, Copilot and Grok. Check the Referral channel for perplexity.ai and other assistant domains, and build a custom channel group if they land there. AI Overviews and AI Mode are excluded from the channel.
Does AI search traffic convert better than organic search?
Often in the published data, but not always and not by a fixed amount. One Seer Interactive client site saw 15.9% vs 1.76% (October 2024 to April 2025); Microsoft Clarity found sign-up rates of 1.66% vs 0.15% across over 1,200 publisher sites (November 2025); Adobe found AI retail traffic 23% less likely to convert in July 2025 and 42% more likely in March 2026. Measure your own rate before you rely on any of them.
How long does it take for GEO to pay back?
There is no independent benchmark, and a fixed payback period quoted in advance is a guess. Set a baseline before you start, measure Layer 1 monthly against control pages, and judge the program on a trend rather than a single month, since AI referrals can swing by half from one month to the next for reasons unrelated to your work.
How does GEO ROI compare to SEO ROI?
No independent study has compared the two per dollar. Run the same formula for each channel with the same cost and margin inputs, using your own conversion rates. Keep them separate: they draw traffic from different sources, convert at different rates and need different decisions.
Can I see AI Overviews and AI Mode traffic separately?
Not as clicks. Google reports those clicks with regular Search traffic, and GA4’s AI Assistant channel excludes them. Search Console’s Generative AI performance reports, available to all websites since August 31, 2026, show impressions, pages, countries and devices for appearances in AI Overviews, AI Mode and generative AI features in Discover.
How much should I budget for GEO?
See How Much Does GEO Cost in 2026? for survey medians and published agency prices, and Best AEO Tools in 2026 for tool pricing. Record the full cost, including internal time, so the ROI figure means something.
Sources:
- Seer Interactive (Nick Haigler and Garman Chan, June 3, 2025). Case Study: 6 Learnings, 1 Site — How Traffic from ChatGPT Converts. One client site, GA4, October 1, 2024 to April 30, 2025.
- Ahrefs (Patrick Stox, June 16, 2025). Does AI Search Traffic Convert Better Than Traditional Search? For Ahrefs, Yes: 0.5% of Visitors Drove 12.1% of Signups. Ahrefs’ own website, 30 days.
- Ahrefs (Louise Linehan, June 24, 2025). AI Visitors Visit Fewer Pages and Bounce More Often Than Traditional Search Visitors. 81,947 sites.
- Microsoft Clarity (Ihab Rizk, November 6, 2025). AI Traffic Converts at 3x the Rate of Other Channels (Study). Over 1,200 publisher and news sites.
- Semrush (Rachel Handley, page dated July 21, 2025). We Studied the Impact of AI Search on SEO Traffic.
- Adobe (Vivek Pandya, August 21, 2025). Generative AI-powered shopping rises with traffic to U.S. retail sites.
- Adobe Digital Insights (April 2026). Quarterly AI Traffic Report (PDF).
- Previsible (David Bell, July 6, 2026). 2026 State of AI Discovery Report. 166 GA4 properties, 6.77 million LLM-referred sessions, November 2024 to May 2026.
- Google Analytics Help. What’s new in Google Analytics (May 13, 2026: AI Assistant traffic measurement) and Default channel group.
- Google Search Central Blog (June 3, 2026). Introducing Search Generative AI performance reports in Search Console.
- Google (Mrinalini Loew, June 3, 2026; updated August 31, 2026). New opportunities, control and insights for website owners.
- Go Fish Digital (Patrick Algrim, September 24, 2025). Generative Engine Optimization (GEO) Case Study: 3X’ing Leads. Agency’s own website, self-reported.
- The Rank Masters (September 12, 2025). Generative Engine Optimization (GEO) Case Study: 8,337% ChatGPT Referral Growth in 90 Days. Agency’s own relaunched website, self-reported.
All sources checked November 10, 2026.
Related reading: How to Measure AI Citations, Best AEO Tools in 2026, How Much Does GEO Cost in 2026?, AI Overviews Cut Clicks by 40% Where They Appear, The Cost of AI Invisibility.
Get a citation audit. We test 30–50 locked prompts across ChatGPT, Google AI Overviews, Perplexity, Gemini, Claude, and the OpenAI Responses API. Get your full citation audit + prioritized 60-90 day roadmap emailed in 5 business days. The audit is $1,500, credited toward your next step, or start with a free 10-prompt Snapshot.