Research · June 2026
WM
Will McGivern-Smith · SVP Partnerships, Fairing · June 2026
Everyone knows AI is changing how people find information. What’s less obvious is how much it’s already changed how people find and buy products — and how invisible that change is in most analytics stacks.
At Fairing, over 30% of our own customers now report discovering us through an LLM. Almost none of that shows up in Google Analytics. The referrer is missing, the UTM is blank, and the session looks direct. The customer arrived through AI and left no trace in the tools we’ve spent years trusting.
We suspected our brands were seeing the same thing. So we went looking for proof.
We analyzed 104 weeks of data across 158 ecommerce brands doing between $10M and $2B in annual GMV — comparing post-purchase survey responses against every click-based signal available: UTM parameters, AI referrer URLs, and any other trackable footprint. The results were larger than we expected.
vs. click attribution
7.5×
more AI-influenced orders captured by post-purchase surveys
vs. click attribution
10.6×
more AI-driven revenue captured by post-purchase surveys
year over year
12×
growth in AI’s share of GMV across the cohort
The channel is real. The measurement just hasn’t caught up. ChatGPT is now the #1 cited discovery source across our customer base.
What % of sales is AI actually driving?
N=158 ecommerce brands · $10M–$2Bn annualized GMV · Weekly, Nov 2024–May 2026
Trend across the full sample, weekly, Nov 2024–May 2026: all three PPS attribution series shown together.
Why Click-Based Attribution Misses AI
The mechanism is structural, not a tracking bug. When someone asks ChatGPT which mattress to buy, gets a recommendation, and navigates directly to your site, there is no click to capture — no UTM parameter, no referrer string. The sale registers as direct, dark, or simply unattributed, regardless of how clean your tracking setup is.
This is the same no-click problem brands have faced with podcasts and influencers for years. AI is the same dynamic at larger scale, accelerating faster, and compounding monthly rather than growing linearly.
Our analysis uses the strongest possible click-based baseline: UTM source matching across all major AI tools plus AI referrer URL patterns. Even so, a 7.5× gap in orders and 10.6× gap in revenue remains.
Category Breakdown: The Numbers Are Not Small
The aggregate numbers mask significant variance by category. Higher-consideration, higher-AOV purchases show the largest AI share. People extensively research a $1,500 mattress or a $400 health device before buying. They don’t ask ChatGPT which gum to buy.
AI % of GMV by category
Feb–May 2026 vs Feb–May 2025 · extrapolated via post-purchase survey
Feb–May 2026
Feb–May 2025
Winners and Losers Are Emerging — Fast
The most significant finding isn’t the category averages. It’s the spread within categories, and how quickly it’s widening. A year ago, the gap between the top and bottom AI-attributed brand in Housewares was about 2 percentage points. Today it’s over 20 points. Every single category shows the same pattern: the spread is widening dramatically.
This isn’t random variance. Brands at the top are building real presence in AI recommendation outputs. Brands at the bottom aren’t necessarily losing AI-driven customers — they’re just blind to them, unable to measure, attribute, or act on the signal. The gap is compounding every month.
Leaders vs. median: estimated annual AI revenue premium
Feb–May 2026 · Top brand vs. category median · Annualized based on average brand GMV per category
Top brand AI % of GMV
Category median AI % of GMV
Laggards vs. median: estimated annual AI revenue missed
Feb–May 2026 · Bottom brand vs. category median · Annualized based on average brand GMV per category
Bottom brand AI % of GMV
Category median AI % of GMV
What This Means
The measurement gap is real, but it’s addressable. Post-purchase surveys that explicitly ask customers whether an AI tool influenced their purchase are currently the most reliable way to surface this signal — not because they’re perfect, but because click-based attribution was simply never designed for a no-click channel.
Beyond measurement, understanding your AI visibility matters. Tools like Profound can help brands see how and where they appear in AI recommendation outputs. Comparing your numbers against category benchmarks — like those in this analysis — gives context for whether you’re ahead or behind.
For brands looking to improve their AI presence, Cloudflare’s isitAgentReady is a useful starting point for understanding how AI-accessible your site is. Most major LLMs also draw heavily from user-generated content — Reddit threads, G2 reviews, Trustpilot, and similar platforms. Brand presence and positive sentiment on those sites increasingly feeds AI recommendations directly. It’s less a new strategy than an extension of what good brands are already doing.
The competitive gap is a harder problem. Brands that can measure the channel are better positioned to act on it. Brands that can’t are making decisions with an incomplete picture.
A full category-level breakdown and deeper methodology is coming shortly.
Methodology & caveats
158 brands with $10M–$2B annualized GMV, Nov 2024–May 2026 (104 weeks). Brands must run an AI option on their HDYHAU survey and have ≥50% order-level UTM capture.
Click attribution = UTM source matching (chatgpt%, perplexity%, claude%, gemini%, copilot%, grok%, etc.) plus AI referrer URL patterns — the strongest available click-based baseline.
PPS extrapolation assumes survey respondents represent non-respondents. Response rates vary by brand.
Dollar gap estimates use average brand GMV within each category, annualized from the Feb–May 2026 period. Individual brand results will vary.
Multi-touch attribution is hard. AI-influenced orders often involve other touchpoints. We measure incremental lift over click-based methods, not sole attribution.
