What attribution actually is
Attribution is the work of deciding which marketing channel deserves credit for a conversion. A user touches three or four channels on the way to becoming a lead or making a purchase. Attribution rules distribute the credit across those channels. The simple version (last-click) gives all credit to the last touch. The complicated version (data-driven, time-decay, linear) spreads credit across the path.
For SEO specifically, attribution matters because organic search rarely sits in the last-click position. The user finds you through organic, leaves, thinks about it, comes back via a direct visit or branded search, and converts. Last-click attribution gives all the credit to the direct or branded visit and none to the original organic discovery. Data-driven attribution spreads the credit more honestly but harder to defend at a board level.
None of the attribution models are correct. They are all simplifications of a multi-touch reality. The job is to pick the simplification that is least wrong for your business and to be honest about its limits.
Why SEO attribution is messy
Four structural reasons SEO attribution is harder than paid attribution.
Reason 1: SEO is rarely the last touch
SEO usually plays an early-to-middle-funnel role. A user searches "how to fix X", lands on your blog, learns about your service, leaves, and converts a week later through a direct visit or a branded search. Last-click attribution gives the direct or branded visit all the credit. The original organic touch gets ignored. The result is that SEO is systematically under-credited under last-click, particularly for service businesses with longer consideration cycles.
Reason 2: Dark traffic is real and growing
A meaningful share of organic-equivalent traffic now lands in GA4 as "direct" because the referrer header was stripped. Privacy-protected browsers (Safari with Intelligent Tracking Prevention, Firefox with Enhanced Tracking Protection, Brave). In-app webviews (Instagram, Facebook, LinkedIn, TikTok in-app browsers). Email apps that strip referrers when the user clicks a link. AI search clients that drop the referrer. The dark traffic share has grown noticeably since 2023 and now sits at meaningful levels across most client GA4 properties.
Reason 3: Branded versus non-branded ambiguity
Branded search is technically organic, but the demand was created by something else (offline marketing, PR, paid media, word of mouth). Attributing branded organic to "SEO" inflates the SEO credit; attributing it to the offline driver is more honest but harder to defend without a clean tracking link. The honest split treats branded organic and non-branded organic as different channels for attribution purposes. See the GSC chapter for the regex pattern that separates them.
Reason 4: AI search referrers
ChatGPT, Perplexity, Gemini and Copilot pass through referral data inconsistently. Some hits land cleanly with the AI domain as the referrer; others land as Direct because the client strips the referrer at the network layer. The honest framing accepts that AI-referred traffic is partially visible and likely under-reports the true AI search contribution. See tracking AI referrals.
The GA4 attribution models
GA4 supports six attribution models. The two that matter for most SMBs are data-driven and last-click; the others are mostly academic.
Data-driven attribution
GA4's default. Uses machine learning to distribute conversion credit across the channels a user touched, based on observed patterns in users who did and did not convert. Theoretically more accurate than rule-based models because it learns from your actual data. Practically harder to audit because the credit allocation is opaque.
Strengths: captures the multi-touch reality, gives SEO credit for assist-channel roles, adjusts to your specific conversion paths. Weaknesses: hard to explain in a meeting, can shift unexpectedly if conversion volume is low, only available for properties with enough data.
Last-click attribution
Old-school but defensible. Gives full credit to the last channel that touched the user before they converted. Simple, easy to explain, audit-friendly.
Strengths: easy to defend in a board meeting, easy to reproduce in a spreadsheet, consistent over time. Weaknesses: systematically under-credits early-funnel channels like SEO, over-credits bottom-funnel channels like direct and branded search.
Our default usage
We use data-driven for internal analysis (the truer multi-touch picture) and last-click for stakeholder reporting (the explainable picture). When pressed on the SEO contribution, we frame it as a band between the two: "Last-click says SEO drove 35 percent of leads, data-driven says 50 percent. The honest band is 35 to 50 percent." The band is defensible; arguing for a single point estimate is not.
See GA4 for SEO for the attribution-setting walkthrough inside GA4.
Dark traffic and the direct bucket
The Direct channel in GA4 includes three different things mashed together. Users who typed the URL or used a bookmark (real direct traffic). Users who came from a channel that stripped the referrer (functionally organic, paid, social or email but attributed wrong). Users who reset their session after the original referrer expired (returners with no fresh referrer).
The honest way to handle Direct in SEO attribution. Look at the landing-page distribution. Real direct traffic concentrates on the homepage and a few well-known internal pages. Dark traffic distributes more broadly across deep landing pages (blog posts, service-area pages, niche product pages) that nobody types directly into the browser. Treat direct traffic to deep landing pages as functionally organic for attribution analysis.
The estimate we publish to clients is a band, not a number. "Direct traffic to deep landing pages accounts for roughly 30 to 50 percent of the total direct count, and is functionally organic." We document the assumption, show the working, and let stakeholders challenge it. Documented assumptions are more credible than pretending the Direct number is clean.
AI search referrers
The AI search referrer picture sits inside the broader attribution honesty conversation. Three patterns matter.
Pattern 1: Inconsistent passthrough. ChatGPT sometimes passes referrer data, sometimes does not. Same for Perplexity, Gemini and Copilot. The share of AI-referred traffic that lands cleanly in GA4 with the AI domain as the referrer is partial and changes as the vendors update their clients.
Pattern 2: AI traffic has unusually high conversion intent. The user has already asked a specific question and is clicking through to verify or buy. Conversion rates on AI-referred sessions often run higher than on regular organic sessions across the client base we measure, which makes the under-counting more painful.
Pattern 3: The cleanest fix is a custom channel. Build a GA4 custom channel grouping that catches the known AI referrer hostnames (chat.openai.com, chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com) and surfaces them as AI Referrers. Accept that some AI-referred traffic is still hiding in Direct, and frame the AI contribution as a partial-visibility band. See tracking AI referrals for the exact build.
The honest framing for stakeholders
The framing that survives a board meeting. Three principles.
Principle 1: Report attribution as a band, not a point. "SEO drove 35 to 50 percent of qualified leads last quarter" is more defensible than "SEO drove 42.7 percent of qualified leads last quarter". The band signals that the underlying measurement has known limits; the point estimate signals false precision.
Principle 2: Document the assumptions out loud. Stakeholders trust honest measurement more than precise-sounding measurement. Write down the assumptions (which referrers count as organic, how dark traffic is being treated, which attribution model is being used for the report) and put them on the same page as the numbers. Anyone who wants to challenge them can; most will not, but they will respect that the option exists.
Principle 3: Cross-reference with CRM source data where possible. Ask sales how each lead heard about the business at intake. CRM source data is imperfect (users misremember, sales staff abbreviate), but it provides a parallel measurement that catches gaps in the GA4 picture. The intersection of GA4 attribution, CRM source data, and call-tracking data is more defensible than any one source alone. See reporting to stakeholders for how to surface this in the monthly report.
The wider point. Honest attribution is a competitive advantage. The agencies that lose client trust over attribution are the ones that reported precise numbers as if they were ground truth, then could not defend them when scrutinised. The agencies that keep client trust report defensible bands, document the limits, and treat the attribution conversation as ongoing rather than as a settled question.
Common mistakes
- Picking an attribution model and sticking with it for trend continuity.
- Documenting the dark-traffic assumption out loud and putting it in the report.
- Building a custom AI Referrers channel in GA4.
- Reporting the SEO contribution as a defensible band (last-click to data-driven range).
- Cross-referencing GA4 attribution with CRM source data and call-tracking.
- Switching attribution models mid-quarter and breaking trend continuity.
- Pretending the Direct channel is clean.
- Reporting a single precise SEO contribution number without acknowledging the limits.
- Letting branded organic credit hide the absence of non-branded growth.
- Treating AI-referred traffic as too small to bother with.
Perth and WA context
Two attribution patterns specific to Perth and WA businesses.
Phone leads under-count organic in GA4. Service businesses across Perth (trades, healthcare, legal) take a large share of leads by phone. Without call tracking (CallRail, WhatConverts, Ringba) the phone-call attribution defaults to whatever the user's last session source was, which is often Direct or branded. The honest fix is to layer call tracking on top of GA4 so the call records the original GA4 client ID and the call attribution flows back to the right channel. See trades SEO and legal SEO.
Local pack clicks attribute weakly. Users who click a phone number directly from the local pack on mobile sometimes never visit the website at all. The lead never lands in GA4. For these users, attribution lives in Google Business Profile insights and the call-tracking data, not in GA4. Treat GBP insights as a parallel attribution source for local businesses. See Google Business Profile and Local SEO Perth.
For the wider context, the GA4 chapter covers the attribution model settings inside GA4, the SEO KPIs chapter covers how attribution feeds the primary KPI, the reporting chapter covers how to present attribution honestly in stakeholder reports, the SEO vs SEM vs paid chapter covers the channel-mix attribution question more broadly, the search intent chapter covers the intent segmentation that improves attribution quality, and the tracking AI referrals chapter covers the AI-specific attribution build. Clients ready to build a more honest attribution stack can engage the SEO service or start with a free SEO audit.