After ChatGPT's 7 May 2026 update, Similarweb found that about 60% of ChatGPT referrals landed on homepages, up from roughly 26–32% before. (Similarweb) Your analytics will file most of those visits as brand demand, and that's one more reason most B2B SEO reports answer the wrong question. They show traffic and rankings. The CFO wants to know how much pipeline came from organic search, and whether to spend more.
That question has always been hard in B2B. Sales cycles are long, buying committees are big, and the person who reads your comparison page is often not the person who books the demo.
AI search has made it harder. A buyer can read your name in an AI assistant's answer, never click, and type your URL into a browser three weeks later. In your analytics, that's "direct" traffic.
Our position: stop trying to make last-touch analytics tell the whole story. It can't, and the harder you push it, the more organic and AI search will look like they do nothing. Ask buyers where they heard about you, and treat that answer as a primary source. This guide covers why attribution breaks, a four-layer setup that's honest about what each layer can and can't see, a sample report, and how to present it to people who sign budgets.
Why B2B SEO attribution breaks
Four reasons stand out, and AI search makes two of them worse.
1. Long, multi-person journeys
Gartner's survey of 645 B2B buyers (August–September 2025) found buyers used an average of seven information sources during a recent purchase. (Gartner) A last-click model gives all the credit to whichever of those seven came last, which is often a branded search or a sales email.
2. Brand and non-brand get mixed up
When someone searches your brand name, SEO didn't create that demand. A non-brand search that ends on your comparison page is the demand SEO actually earned. Many reports lump the two together.
3. AI answers don't always send a click
Semrush's "ghost citations" study found in 61.7% of AI citations the answer never named the brand, and about a quarter of appearances named a brand without linking it. (Semrush) Either way, there's often nothing for your analytics to record.
4. AI clicks that do arrive look like something else
That homepage shift is only half of it. Similarweb also notes that traffic filed as "direct, no clear source" may increasingly be AI-driven discovery without a clean referral tag. (Similarweb)
So a homepage visit from ChatGPT looks a lot like brand demand, and some AI influence looks like direct traffic. If you only read referral data, you'll undercount AI.
The four-layer setup
No single tool sees the whole journey. The fix is to stack four sources, each covering a gap in the others.
Layer 1: Search Console, split by brand and AI features
Search Console now does two things that used to need workarounds.
Branded vs non-branded queries. Google introduced a branded queries filter in November 2025 and made it available to all eligible sites on March 11, 2026. It's only available for top-level properties with enough query volume. (Google Search Central Blog) Google notes some queries may be misidentified, and the filter has no effect on ranking.
Generative AI performance. Google's Generative AI performance report (June 2026) shows impressions from AI Overviews and AI Mode, with page, country, device and date breakdowns. (Google Search Central Blog)
What it tells you: whether non-brand visibility is growing, and how much of it comes through Google's AI features. What it can't tell you: anything about pipeline, or about ChatGPT and other assistants.
Layer 2: Web analytics, with AI as its own channel
GA4 now has an AI Assistant default channel for visits from sources such as ChatGPT, Gemini, DeepSeek, Copilot and Grok. Note the exception: it excludes Google's AI Overviews and AI Mode. (Google Analytics Help) Those clicks show up as organic search.
Set up landing-page reports for organic search and AI Assistant separately, and watch homepage traffic from both.
What it tells you: which pages and channels bring visitors, and which of those visitors convert on-site. What it can't tell you: influence without a click, or what happens after a lead hands off to sales.
Layer 3: CRM, with first touch and self-reported attribution
This is where pipeline lives, so it's the layer that matters most.
Pass the first-touch source and landing page into the CRM with every form fill, using hidden fields. Then you can report pipeline by original channel, not just by last touch.
Ask "How did you hear about us?" on demo and signup forms, with AI tools (ChatGPT, Perplexity, Gemini, Claude) listed as options alongside Google search. Leave a free-text field too.
Self-reported data catches what analytics misses. Ahrefs asked exactly this and learned that 21.7% of its free and paid signups credit Claude or ChatGPT, ahead of YouTube and behind only Google. (Ahrefs) That's one company, with a very AI-savvy audience, so don't expect your number to match. But it shows how much a referral report can miss.
The catch is timing. A buyer who first heard of you in a ChatGPT answer may not fill in your demo form for a quarter or more, so this field pays off slowly, and you need to start collecting it long before anyone asks for the report.
What it tells you: which channels sourced and influenced real opportunities, in buyers' own words. What it can't tell you: much in the first few months, because of that delay. And people misremember, so treat it as a strong signal, not an exact count.
Layer 4: AI visibility tracking
Track a fixed set of buyer prompts in the AI assistants your buyers use. Log whether you're mentioned, cited and described accurately. Tools from Ahrefs, Semrush and Similarweb can automate this, or you can run a smaller set by hand.
Google's optimization guide adds a caution: "No third-party tool has access to our internal ranking or AI systems." (Google Search Central) These tools sample answers. They don't see inside the engines.
What it tells you: whether you're in the answers that shape shortlists. It's a leading indicator. What it can't tell you: revenue. A mention is a mention until a deal shows up in layer 3.
For more on how answer engines choose sources, see What is AEO?. For how AEO and SEO metrics differ, see AEO vs SEO.
Putting it together: a sample monthly report
This is the shape of a report worth sending a B2B SaaS leadership team. It is a template with no figures in it: fill in your own numbers. Nothing in it is real data.
Section | Metric | Source | Why it's here |
|---|---|---|---|
Pipeline | Opportunities and pipeline value with organic or AI search as first touch | CRM | The number the business cares about |
Opportunities where buyer self-reported Google or an AI tool | CRM form field | Catches influence analytics can't see | |
Demos and SQLs from organic and AI Assistant channels | CRM + GA4 | Shorter-term conversion signal | |
Demand | Non-branded clicks and impressions | Search Console | Is SEO creating new demand? |
Branded clicks and impressions | Search Console | Is overall brand demand growing? | |
AI Overviews and AI Mode impressions | Search Console Gen AI report | Visibility in Google's AI features | |
AI visibility | Share of tracked prompts where you're mentioned, cited, accurate | Prompt tracking | Leading indicator for AI shortlists |
Pages | Top landing pages by demos (not sessions) | GA4 + CRM | Which content earns conversations |
Put pipeline first. If leadership only reads one row, it should be that one.
How to present SEO attribution to a CFO
Show ranges, not false precision. "Organic and AI search were the first touch on 18–25% of new pipeline this quarter, depending on how we count self-reported answers" is more credible than a single exact figure. (Illustrative placeholder only: 18–25% is a made-up figure to show the format. It is not real data, a client result or a benchmark.)
Separate sourced from influenced. Sourced means organic was the first touch. Influenced means a buyer touched organic content somewhere in the journey. Report both, label them clearly, and don't add them together.
Expect a lag. In B2B, a page published this quarter may not show up in closed revenue for two or three. Say so before the first report, not after.
Don't hide branded search. Branded demand isn't SEO's achievement alone, but losing it would hurt. Report it separately so nobody mistakes it for non-brand growth.
Record mentions as mentions. Miraa's AEO page puts it this way: "A mention with no business result is recorded as a mention." (Miraa) Visibility is worth tracking. It just isn't pipeline.
Common attribution mistakes
Reporting sessions as success. Traffic from glossary posts can grow while pipeline stays flat. Miraa's B2B SaaS page calls a traffic chart that never meets the language of demos and pipeline "the stall, even when the chart goes up." (Miraa)
Trusting last click. In B2B, last click usually credits branded search or direct, which hides the content that created the demand.
Treating AI referrals as the whole AI story. As the Similarweb and Semrush data above shows, a lot of AI influence never shows up as a referral.
Changing the method every quarter. Pick definitions for sourced, influenced and self-reported, write them down, and keep them stable so trends mean something.
Ignoring data quality in the CRM. If sales reps overwrite the lead source or forms drop hidden fields, every report built on top is wrong. Audit a sample of records each quarter.
A 30-day setup checklist
[ ] Turn on the branded queries filter and check the Generative AI report in Search Console.
[ ] Confirm the AI Assistant channel appears in GA4 and build landing-page reports for it and for organic search.
[ ] Add hidden fields for first-touch source and landing page to every form, and map them to the CRM.
[ ] Add "How did you hear about us?" with AI tools as options, plus free text.
[ ] Write down definitions for sourced, influenced and self-reported pipeline.
[ ] Choose 20–50 buyer prompts and record a baseline across the AI assistants your buyers use.
[ ] Build the monthly report with pipeline in the first section.
Final thoughts
Perfect SEO attribution in B2B doesn't exist, and AI search moved it further away. That's fine. The goal is a method that's honest about what it can see, stable enough to show trends, and tied to pipeline instead of pageviews.
Put self-reported attribution in place this month, because it takes the longest to pay off. Then tighten the CRM and analytics layers around it. To decide which pages to build so there's pipeline to attribute, see our B2B SaaS SEO guide. Don't wait for a perfect model before you report; a range you can defend beats a single number nobody believes.
If you're evaluating agencies, ask each one to show you a real report. The roundup of B2B SaaS SEO agencies compares agencies on pipeline proof. Miraa reports demos, SQLs and pipeline sourced from organic and AI search. Its B2B SaaS SEO page describes that reporting.