Can You Actually Measure the ROI on SEO and AEO?

Both SEO and AEO are measurable, but not in the way most businesses are currently set up to measure them. AI is answering your buyers before they reach your website, which means traffic is no longer a reliable proxy for commercial influence. The frame that works is pipeline value: who converted, from where, and what the deal was worth. Getting there means connecting your content to your CRM, understanding what your attribution tools can and cannot see, and adding the signals that live outside them.

Here is the practical version of the problem. Your CFO asks what the SEO retainer returned last quarter. You open GA4, organic sessions are down fourteen percent, and you know the honest answer is "more than that number suggests" without being able to show the working. That gap, between what you are reasonably confident is happening and what you can put in front of a board, is what this article is about.

In B2B, where buying cycles are long and decisions involve several people researching across several channels, conflating activity with commercial outcome is one of the most expensive measurement mistakes available to you.

How SEO measurement works, and where it breaks down

Traditional SEO has a relatively clear line from search engine to commercial result. Someone searches for something relevant, your content appears, they click, they land on your site, and at some point they take an action. Google Search Console tells you what they searched and how often they clicked, and its Generative AI performance report shows separately how often your pages turned up inside AI Overviews and AI Mode. GA4 tells you what they did when they arrived. Your CRM tells you whether any of them became a lead, a qualified opportunity, or a customer.

The standard calculation is straightforward: take the pipeline value or revenue attributable to organic search, subtract what you spent on SEO (agency fees, content production, tooling), and divide the result by that cost. The harder part is making sure your CRM is configured to capture the full journey rather than just the first or last touchpoint.

Two reports make that problem visible. The first is a journey path view, which shows how often a single buyer comes back through a different channel each time.

The second is a path level attribution report, which shows the sequences that actually preceded closed revenue, and how rarely any of that revenue belongs to a single channel.

Where B2B SEO measurement typically fails is at the attribution layer. Most businesses know which contacts came from organic search at the point of form submission and credit that last touch, without accounting for the long run of interactions the contact moved through before arriving there. You will never pin the exact contribution down, and you do not need to. What you can do is model it: assign weighted credit by channel, then watch how that weighting shifts as deals close or stall. A model you review every quarter will tell you more than a precise number you cannot produce.

The AEO problem: your buyer is already being answered before they reach you

AEO presents a fundamentally different measurement challenge, and pretending otherwise will lead you to underinvest in it or misread its contribution.

When a buyer types a question into ChatGPT, Perplexity, or Google's AI Overview, they often receive a direct answer without visiting any of the sources cited. If your brand is mentioned in that answer, you have influenced their thinking. If a competitor is mentioned instead, you have lost ground. Neither of those events shows up in your traffic data, your form submissions, or your pipeline report. They simply happen, quietly, while your dashboards tell you nothing. In the first four months of 2026, 68 percent of US Google searches ended without a click. Two years earlier it was 60 percent. Pew Research, which watched 900 US adults browse, found the same pattern from the other side: people clicked a search result 8 percent of the time when an AI summary appeared, and 15 percent of the time when it did not. We have written separately about what this shift means for the future of SEO.

This is sometimes called the dark funnel, though a more useful frame is this: the buyer's research process has become partially invisible. That does not mean you cannot measure AEO's contribution. It means you have to measure it differently, using four signals rather than a traffic number. If the distinction between SEO, AEO and GEO is still blurry, our guide to all three sets out where each one applies.

1. The first and most directly trackable signal is referral traffic from AI platforms. Analytics tools are evolving fast and are already beginning to surface this traffic natively: some platforms label it AI Referrals, others AI Assistant. When a buyer does click a citation link from ChatGPT, Perplexity, or a similar engine, that visit is attributable. This is a real and growing cohort, but the probability of that click is low, and it underrepresents total AI influence by a significant margin. Remember that the objective is not traffic to your website. Much of what a buyer wants is being gathered outside it.

It can also mislead in the other direction. On Semrush's data, AI referral traffic converts at roughly 4.4 times the rate of standard organic. Before you take that as proof the channel is working, understand what sits behind it: much of that traffic is navigational. The buyer had already decided and asked the assistant for the link. A strong conversion rate tells you the traffic is late stage. It does not tell you that AEO created the opportunity.

2. The second signal is brand visibility. This is how often your brand appears in AI-generated answers across a set of prompts you choose to track. HubSpot's AEO dashboard scores it on prompt coverage, consistency across engines, and how often competitors turn up in the same results, running your prompts daily across ChatGPT, Gemini and Perplexity. Visibility can be actively influenced through AEO work: structuring your website and content so that AI engines can read, index, and draw from it with confidence. You can now pay for placement too. OpenAI's ChatGPT Ads Manager is open to advertisers, and it buys you a sponsored card below the answer, clearly marked as an ad. It does not buy you a place inside the answer. OpenAI says its ads run on separate systems and do not change what the model tells people.

The complication is that visibility is not a single number. Your brand sits differently in ChatGPT, Perplexity, Gemini, Claude and Copilot, because each one draws on different sources and weights them differently. Being the first recommendation in one engine and absent from another is normal, and an average across all of them hides exactly the gap you need to act on. Tracking brand visibility means tracking it in each engine separately, not just in the one with the largest user base today.

3. The third signal is AI share of voice. Visibility tells you whether you appear. Share of voice tells you how much of the conversation belongs to you. HubSpot calculates it by dividing your brand's mentions by the total brand mentions from all tracked competitors across the same prompts, and reports it under the competitor landscape section of the AEO dashboard. The two numbers can move in opposite directions, which is the reason to keep both. A brand can appear in most answers and still hold a small share of the mentions inside them, meaning it is present but not preferred, and that is a different problem requiring a different fix.

Read share of voice with two caveats. The first is that it is bounded by your prompt set. HubSpot tracks 25 prompts a day on Marketing Hub Professional and 50 on Enterprise, so the figure describes your position on the questions you chose, not the market. Choose the questions your sales team actually hears, not the ones that flatter you. The second is that answer engines are not deterministic. Ask the same engine the same question twice and you will often get a different answer, which means a dashboard reading 17.4 percent, up two places this week, is more precise than the data behind it. Report direction over a quarter rather than the decimal, and pick a tool that lets you read the raw answers sitting behind the score.

4. The fourth signal is mentions and citations, and it is arguably the most important of the four. A mention is when an AI engine references your brand by name in a generated answer. A citation is when it links back to your content as a source. The distinction matters: citations signal that an AI system considers your content authoritative enough to surface as evidence, not just relevant enough to name. They are also partially trackable through referral data when a user follows the link. The two are genuinely separate, and a source can be cited without mentioning your brand at all. Monitoring how frequently your brand is cited, which content earns citations, and whether your citation share is growing or declining relative to competitors tells you something no traffic report can: whether AI engines trust you as a source. This is the foundation of AEO. Everything else builds on it.

For Google's surfaces there is one figure you do not have to buy. Search Console's Generative AI performance report counts how often your pages appeared inside AI Overviews and AI Mode, so unlike a sampling tool it gives you a count rather than an estimate. It stops short of clicks, click-through rate and queries, so read it as a visibility number and not as a measure of return.

Measuring both together: the attribution model that works in B2B

The frame for measuring SEO and AEO return should be pipeline value, not traffic volume, and this distinction matters more now than it ever has. AI is actively moving your content outside your trackable territory. By the time a buyer lands on your website, most of their questions have already been answered. They arrive ready to act, not ready to browse. That means visitor numbers may drop. That is not failure. If your strategy is working, the number of deals will increase.

The deeper issue is that last-touch attribution is structurally inadequate for B2B buying behavior. Forrester's B2B buying study puts the average significant purchase at 27 buying interactions, up from 17 two years earlier, split roughly between 15 digital and 12 human. Sixty percent of those purchases now involve four or more people in the buying group, up from 47 percent in 2017. And 32 percent of purchase journeys run four months or longer, against 19 percent in 2015.

Read those three numbers together and the problem with last-touch attribution becomes arithmetic rather than theoretical. You are assigning the full commercial value of a four-month decision, made by four or more people across 27 interactions, to whichever link happened to be in the browser when someone filled in a form. No single channel closes a deal in that environment.

What you need instead is a cross-channel attribution model that assigns weighted relevance to each channel based on its actual role at each stage of the decision journey. A data-driven model, increasingly supported by AI-assisted analysis, distributes credit proportionally across every channel that contributed to a conversion, weighting each by how often it turns up in journeys that ended in a sale compared with ones that did not. This is a more honest and commercially useful picture than any single metric will give you. It still cannot tell you whether those channels caused the sale or were simply present for it.

Setting this up in HubSpot

If HubSpot is your CRM, you have most of what you need already. The configuration work is specific and not particularly complex, but it requires intention. It does depend on your underlying data being structured properly in the first place, which is a first-party data problem before it is a reporting one.

HubSpot offers two types of pre-built attribution reports that serve different purposes. The first measures impact by contact, showing which sources, content assets and interactions had the greatest influence on generating leads. The second, available in the Advanced Marketing Reports, measures impact by deal, connecting touchpoint performance to revenue through customer journey and multi-touch attribution reports. Used together, they give you visibility across both the lead generation layer and the commercial outcome layer, which is the combination most businesses are missing.

HubSpot classifies visits from AI assistants under its own AI Referrals traffic source, and names the platform in traffic source drill-down 1. Three reports are worth building: AI-referred sessions by platform, contacts whose Original Source is AI Referrals, and deals with an AI Referrals touch anywhere in the path.

Treat that number as a floor rather than a measurement. HubSpot matches a fixed list of assistant domains, including chatgpt.com, claude.ai, perplexity.ai, copilot.microsoft.com and gemini.google.com. Anything from an engine not on the list, or from a buyer who copied the URL out of a chat window rather than clicking it, lands in Direct instead. The AI Referrals number is real. It is simply smaller than the influence it represents, and the gap between the two is not something you can close with better configuration.

HubSpot's multi-touch attribution is powerful within the boundaries of what it has actually tracked. It can look across a contact's full recorded journey, applying different models to show which channels contributed to a deal rather than just which one closed it. That is genuinely useful for SEO measurement and necessary for understanding relationships that build over long cycles. But it cannot infer what it never saw. It works from the first tracked session forward, meaning anything that happened before that contact appeared in your system is invisible to it. The dark funnel does not become visible. HubSpot tells you an accurate story about what it can see. The problem is that in an AI-influenced buying journey, a significant part of the story happens before anyone arrives.

The one signal your CRM cannot produce: ask the buyer

There is one source of evidence that reaches into the part of the journey no system can reconstruct, and it costs nothing to add. Ask the buyer directly.

A short open-text question on your demo and contact forms, asking how they first heard about you, picks up the conversations that happened in ChatGPT, in a private Slack channel, or at an industry event. Keep it as free text rather than a dropdown, because a dropdown tells people which answers you were expecting and most will pick the nearest one rather than the true one.

The work is in what happens next. Someone needs to read and sort those responses each week into a consistent set of categories, with AI assistants as one of them, and store the result against deal value. That is a standing responsibility, not a one-off setup task, and it is the step most businesses skip. Buyers also misremember, so give it two quarters before you read a trend into it. Even allowing for all of that, it is the only view you will get of influence that never produced a click.

The bigger commercial risk is treating zero-click invisibility as invisible results. If your competitors are being cited in AI answers and you are not, that gap is compounding over time. You are not losing traffic. You are losing the recommendations.

Where this leaves you

Search is still doing the work. What has changed is where the evidence sits. Organic sessions, new visitors, keyword position and click-through rate were always proxies for influence, and in a zero-click search market they are going to keep falling. Defending them puts you in an argument you cannot win.

Reporting that holds up in front of a board has three parts: 1) attribution that runs through to deal value rather than stopping at the form, 2) visibility tracked engine by engine rather than averaged into one number, and 3) an open-text source question on your forms that someone reads every week. 

Working out what your search and AI visibility is actually worth is the part most teams get stuck on. It is also most of what our B2B SEO and AEO work involves: connecting visibility to pipeline rather than to sessions. If you want a clear view of what your reporting captures today and what it is missing, book a consultation.

FAQs

How do I make my website readable and citable by AI engines?

There are two layers to this. The first is technical access: check that your robots.txt file is not blocking AI crawlers such as GPTBot, ClaudeBot or PerplexityBot, and make sure your pages load cleanly with structured markup. The second is authority: write direct answers to the questions your buyers are actually asking, stay focused on a defined set of topics, and earn references from credible third-party sources. AI engines surface content they consider trustworthy. You cannot control which answers they include you in, but you can control how easy you make it to find you, read you, and trust what you say.

How long before AEO work shows up in pipeline?

Visibility moves faster than pipeline. Citations and mentions in AI answers can shift within weeks of restructuring content, because the engines reindex continuously rather than on a crawl schedule you can predict. Pipeline takes as long as your sales cycle takes, which for most industrial and professional services businesses means two to three quarters before the effect is legible in closed revenue. The practical consequence is that you need two reporting layers running at different speeds: a visibility layer you review monthly to confirm the work is landing, and a pipeline layer you review quarterly to confirm it is worth paying for. Judging AEO on pipeline alone in the first six months will tell you it has failed when it has not yet had time to.

What is the difference between a mention and a citation in AEO?

A mention is when an AI engine names your brand in a generated answer. A citation is when it links to your content as a source. Citations carry more weight because they signal that the engine treats your content as authoritative, not just relevant. They are also partially trackable through referral data when a user follows the link, and the two are genuinely separate: a source can be cited without mentioning your brand at all. If you are monitoring one metric to understand your AEO position, citation share is the one that tells you whether AI engines trust you as a source.

What is the difference between brand visibility and AI share of voice?

Brand visibility is how often you appear across the prompts you track. Share of voice is your proportion of all brand mentions in those same answers, competitors included. A brand can score well on visibility and poorly on share of voice, which means it is turning up in the conversation but losing it. Both are bounded by the prompt set you chose, so the numbers describe your position on those questions rather than the market as a whole. Use visibility to check whether you are present and share of voice to check whether you are preferred.

My website traffic is dropping. Does that mean my strategy is not working?

Not necessarily, and in an AI-influenced market it may mean the opposite. AI engines are answering buyers' research questions before those buyers ever visit your site. By the time someone lands on your page, most of their questions have already been answered and they are closer to a decision. If your AEO is working, fewer people will arrive mid-research, but the people who do arrive will be more ready to act. The number to watch is not sessions. It is deals. Traffic can also fall for reasons that have nothing to do with AI, which is worth ruling out first: here is how to tell an algorithm update from an AI effect.

Can I build a HubSpot report that shows AI channel performance?

Yes, and in two places rather than one. For traffic, HubSpot classifies AI assistant visits under its AI Referrals traffic source, which carries through to Original and Latest Traffic Source on the contact, so you can tie it to deals and revenue. For visibility, HubSpot AEO tracks brand visibility, share of voice and citations across ChatGPT, Gemini and Perplexity. Between them they cover both halves: the buyers who clicked, and the ones who only saw your name in an answer. What neither can do is join the two, which is what a "how did you hear about us" field on your forms is for.

Where do I start if I want to measure AI's influence on my pipeline?

Start with the two things that cost nothing and take a week: add the open-text "how did you hear about us" field to every high-intent form, and open the Generative AI performance report in Search Console to see which pages Google is already pulling into AI answers. Then move to your attribution reports and connect lead sources to deal-level revenue, so you can see which channels generate commercial conversations rather than traffic. Then look at your referral traffic and identify visits coming from AI platform domains. Only then pay for a tool that tracks visibility and citation share across the major engines, because the earlier steps will tell you whether you have a visibility problem worth monitoring. All of it requires someone who owns the data, reviews it consistently, and understands the difference between what is being measured precisely and what is being estimated directionally.

Working out what your search and AI visibility is actually worth is the part most teams get stuck on. It is also most of what our B2B SEO and AEO work involves: connecting visibility to pipeline rather than to sessions. If you want a clearer view of what your reporting captures today and what it is missing, book a consultation.