+40% Traffic and 9 AI Citations a Month: An Ecommerce Case Study After a Technical SEO + GEO Audit
Key takeaways before you read
- Starting point: an outdoor-gear ecommerce store, 800 product pages, 8,500 visits/month, SeAudit score 38/100, zero citations in ChatGPT, Perplexity, or shopping-focused AI Overviews.
- 6 technical and editorial actions rolled out over 10 weeks (February-April 2026).
- Result after 3 months: +40% organic traffic, 9 AI citations/month, SeAudit score 69/100.
- What mattered most: cleaning up crawl budget (indexable facets) + Product/AggregateRating schema accounted for more than half the impact.
The store in question — let's call it TrailCamp to keep it anonymous — sells camping and hiking gear: tents, backpacks, stoves, around 800 active SKUs on a standard ecommerce platform. Founded in 2019, it was pulling 8,500 organic visits a month as of January 2026, a number that had been flat for a year. No mentions in ChatGPT, Perplexity, or shopping-oriented AI Overviews answers, even on queries where the brand had a genuine product edge. The team came in with a simple question: "our products are good, so why are Google and AI engines ignoring us?"
The audit took a morning. Fixing the diagnosis took three months.
The initial diagnosis: score 38/100
The January 2026 SeAudit audit flagged 5 structural problems.
1. Crawl budget swallowed by faceted pages
Every filter combination (color, size, price, brand) generated its own indexable URL. Result: over 14,000 URLs crawled by Googlebot for 800 real products. Crawl budget was being spent on value-less pages, at the expense of new product pages that took weeks to get indexed.
2. Zero structured data on product pages
No Product schema, no Offer, no AggregateRating. For an AI engine trying to answer "which 2-person tent under $300", the absence of structured price, availability, and rating data makes a source invisible next to competitors who do mark up their pages.
3. "Thin" product pages: 90 words on average
A photo, a title, three lines of generic supplier-copy description. No usage context (what weather, what skill level), no comparisons, no answers to purchase objections. Content that says nothing an AI engine could cite.
4. No internal linking between product pages and editorial content
The store's blog (buying guides, comparisons) existed but lived in a complete silo: zero links to product pages, zero links back. Two content worlds that never reinforced each other.
5. No authority signals for AI engines
No llms.txt, no detailed "About" page, no mentions of expertise (product testing, verified reviews highlighted). Nothing that would let an LLM judge the source's trustworthiness before citing it.
The 6 actions rolled out (10 weeks)
Action 1 — Crawl budget cleanup
Canonical tags pointing to the parent product page on all faceted URLs, robots blocking on low-value filter combinations, XML sitemap reworked to surface only the 800 real product URLs. Effort: 3 dev-days. Impact: the strongest of all actions, tied with action 2.
Action 2 — Product + Offer + AggregateRating schema
Rolled out at scale across all 800 product pages via a template: price, availability, average rating, review count. Shopping-focused AI engines (AI Overviews, ChatGPT Shopping) rely directly on these fields to build their comparisons.
Action 3 — Rewriting the 60 most-visited product pages
Went from 90 to 400-550 words: concrete usage context (what weather, what hiker skill level), a technical specs table, a "who it's for" section, and 3-4 Q&As marked up in FAQPage schema.
Action 4 — Bidirectional linking between products and editorial content
Every buying guide now links to relevant product pages, and every product page links back to its matching guide. Result: time on site went up, and AI engines now have a clear semantic map between "advice" and "product."
Action 5 — Structured, highlighted customer reviews
Existing verified reviews (previously buried at the bottom of the page, unmarked) were moved up visually and marked up with Review schema. A direct trust signal, both for Google and for generative engines that readily cite sources with explicit social proof.
Action 6 — llms.txt and a stronger "About" page
Added an llms.txt listing priority guides and categories, and enriched the "About" page with brand history, supplier partnerships, and product certifications.
The results, 3 months later
| Metric | Before (Jan 2026) | After (Apr 2026) |
|---|---|---|
| Monthly organic traffic | 8,500 visits | 11,900 visits (+40%) |
| SeAudit score | 38/100 | 69/100 |
| AI citations/month (ChatGPT, Perplexity, AI Overviews) | 0 | 9 |
| URLs crawled by Googlebot | ~14,000 | ~2,100 |
| Average indexing time for a new page | 18 days | 4 days |
| Pages with at least 1 displayed review | 12% | 78% |
The 9 monthly AI citations came almost exclusively from the 60 rewritten pages and their linked guides — the correlation between "in-depth content + complete schema" and "citation" is direct. None of the 740 untouched pages generated a citation during the period.
Why a "classic" SEO audit wouldn't have been enough
A generalist SEO audit would probably have flagged the crawl budget issue and the thin pages — those are basics. What it would have missed: the direct link between Product/AggregateRating schema and a shopping-focused AI engine's ability to cite a source. An audit focused purely on "Google rankings" measures positions; an SEO + GEO audit also measures whether an AI engine can unambiguously extract a price, a rating, and usage context. On TrailCamp, the 60 rewritten pages gained Google rankings (+2 to +5 positions on average) AND became citable — the two aren't automatically linked; a well-ranked page can still be invisible to a generative AI if it isn't structured for extraction.
What mattered most
Two actions explain most of the result: cleaning up crawl budget (the technical prerequisite without which nothing else is even seen) and Product/AggregateRating schema (the format shopping-focused AI engines consume directly). Rewritten content and internal linking come right behind — without them, AI engines have the data but not the context to choose you over a competitor. llms.txt, as usual, had a marginal effect on its own but amplified the other signals.
Key takeaways
- A crawl budget clogged with faceted pages caps indexing before content even enters the picture.
- Product/Offer/AggregateRating schema is the entry point for shopping-focused AI engines — without it, your prices and reviews are invisible to them.
- 400-550 words with real usage context is enough to turn a "thin" product page into a citable source — you don't need blog-post-level depth.
- Bidirectional product/editorial linking strengthens the site's overall semantic signal.
FAQ
How long before you see results from an ecommerce audit? Expect 8 to 12 weeks for the first measurable traffic effects, and often a bit longer for the first AI citations, which also depend on generative engines' re-indexing cycles.
Do you need to rewrite all 800 product pages at once? No. Prioritize the 5-10% most-visited pages, or those closest to conversion — that's where rewriting effort has the best impact-to-time ratio.
Is Product schema enough without rewriting content? No, it helps with shopping indexing but doesn't create citable context. The two actions are complementary, not interchangeable.
Want to know where your own site stands on these points? Get your score out of 100 in a few minutes, or check out a sample report to see the level of detail. To go deeper on the product angle, our guide to AI-citable product pages walks through the schema method step by step. And if you want the full report with a prioritized action plan, the complete PDF report covers exactly this framework.