This is part one of a living case study. Our own store — the car-care e-commerce site Sniper Design runs — gets recommended in ChatGPT, Perplexity, and other AI tools, even though we’re only now starting deliberate AI SEO work on it. We’re publishing the baseline numbers before optimization, and in a few months we’ll put the after picture next to them. No borrowed study — our own GA4 and Search Console data.
How much AI tools actually send
Over twelve months (July 17, 2025 – July 16, 2026), AI tools brought in 1,867 visits out of a total of 394,000 — roughly 0.5 percent. Nine out of ten AI visits came from ChatGPT (1,688), the rest from Perplexity (126), Gemini (28), Copilot (23), and a handful from Claude and Grok.
Show the data as a table
| Period | Visits |
|---|---|
| Aug 25 | 156 visits |
| Sep | 124 visits |
| Oct | 144 visits |
| Nov | 157 visits |
| Dec | 107 visits |
| Jan 26 | 127 visits |
| Feb | 112 visits |
| Mar | 148 visits |
| Apr | 223 visits |
| May | 241 visits |
| Jun | 171 visits |
Through March the volume held around 110–160 visits a month. In April and May 2026 it jumped to 223–241 — a rise of more than half over the running average. We’re not reading anything into the June dip yet; whether it’s seasonality or noise, the next few months will tell.
Half a percent sounds like a rounding error. But the volume matters less than who those visitors are — and that’s where the data turns.
Fewer visitors, but ones who’ve already decided
A visitor from ChatGPT behaves differently than one who comes from a search engine. They’re not there to “compare options” — they arrive with a specific recommendation the AI assembled for their exact question.
Where AI sends those visitors
Over the year, AI tools dropped people onto 668 different landing pages. That’s a finding in itself: AI doesn’t just cite the homepage — it knows the categories, the articles, the courses, and the wholesale section.
Next to the homepage, the leaders are educational articles and pages that prove the business physically exists — courses, the detailing studio, the brick-and-mortar store. Exactly the content that lets AI systems answer questions like “where can I learn detailing” or “where do I go to buy car care products in person.”
Google AI Overviews: 200,000 impressions in two months
Referrals are only the visible part, though. In 2026, Search Console added a new beta report — Generative AI features — showing how often a site appears directly inside Google’s AI answers (AI Overviews and friends). For our store: 200,000 impressions between May 18 and July 15, 2026, averaging around 3,400 impressions a day.
Two things stand out. First, the volume: a car-care store was visible inside AI answers two hundred thousand times in two months — a layer of visibility GA4 never shows you, because an impression in an AI answer doesn’t produce a visit. Second, the structure: all ten of the most-shown pages are, again, educational articles, not products and not the homepage. The same content that wins classic search feeds the AI answers too.
Why AI recommends a site that never did AI SEO
Our store hasn’t been through a deliberate AI SEO pass yet — that work is just starting. It earned those recommendations with what it has been doing for years for people and for classic search:
- Educational content answering real questions The ten most-visited pages from Google are ten articles (which diesel is highest quality, 100-octane gas, charging a car battery) — not products. Over twelve months they brought in tens of thousands of clicks.
- Structured product copy Products have a benefit-led heading, specs, usage instructions, and a 'what it's good for' section — exactly the blocks AI assembles a recommendation from.
- Categories with a definition and internal links Categories open with a short definition and link out to related guides — the hub-and-spoke pattern we describe in our guide to pillar structure.
- A distinctive brand with physical proof Its own retail store, a detailing studio, training courses, a strong brand voice. An entity AI can remember and verify.
What we’re doing next
That’s exactly why this study is “living.” The baseline is measured; now we’re rolling out on the store the playbook we describe across the individual guides:
- Content for AI — short answers and fact-dense passages in the key categories and articles,
- structured data — Product, FAQPage, and Organization schema in full,
- E-E-A-T signals — authors, hands-on experience, tying the entity to the physical locations,
- measuring AI visibility — a recurring citation test across platforms instead of a one-off impression.
In 3–6 months we’ll add the same charts again — before and after. If the AI share of traffic and the conversion rates go up, you’ll see exactly how much; if they don’t, you’ll see that too. The last-updated date next to the title always tells you how fresh the numbers you’re reading are.
Key takeaways
- AI referrals are a fraction of a percent of traffic today — anyone promising “20 percent of your traffic from ChatGPT within six months” isn’t showing you data.
- The quality, though, is exceptional: ~4× the conversion rate and ~3× the revenue per visit compared with Google. AI sends customers who’ve already decided.
- Visibility in AI is an order of magnitude bigger than referrals: 200,000 AI Overviews impressions in two months — a layer you won’t see in GA4.
- The trend points up — April and May 2026 brought a jump of more than half.
- What gets cited is content, not the catalog: articles, courses, physical locations. An online store with no educational content has nothing to cite.
- Measurable referrals are a floor estimate — part of AI’s influence hides in brand search, direct traffic, and AI answers that never get clicked.
Want to know what AI tools would cite from your site? This site is run by Sniper Design — an AI SEO audit walks your content through the same lens we’re now applying to our own store: what’s citable, what’s missing, and in what order to fix it.
For transparency: the store in this study is Sniper Design’s own e-commerce site — the same agency that runs this site — so we know the data first-hand and publish it with permission. The numbers come from GA4 (Session source dimension, July 17, 2025 – July 16, 2026) and Google Search Console for the same period; impressions in generative AI features come from the beta Search Console report, which has data from May 18, 2026. Conversion rate = session key event rate in GA4. Clicks from Google AI Overviews can’t be separated from regular organic in the data, so the AI referral figures are a floor estimate. Revenue per visit is published as a ratio to Google organic rather than an absolute amount. We’ll update this article with new data once the AI SEO work is in place.
