2026-09-29

The five layers of AI visibility (and which ones you can't DIY)

Every store we audit fails in some combination of the same five layers. Here they are, ordered from an afternoon's work to a genuine engineering program.

1. Let the crawlers in (an afternoon)

Check your robots.txt for GPTBot, ClaudeBot, PerplexityBot and Google-Extended. Blocking them was a defensible choice in 2023; today it's opting out of a sales channel. Publish an llms.txt while you're at it — it's the emerging standard for telling AI systems what your site offers.

2. Make your pages readable (days to weeks)

If your storefront renders client-side, most AI crawlers see an empty HTML shell. Server-side rendering (or pre-rendering) of product pages is the single highest-impact technical fix for JavaScript-heavy stores.

3. Structure your product data (the big one)

Schema.org/JSON-LD markup per product: brand, GTIN, offers with price and availability, aggregate ratings. This is what AI shopping surfaces actually consume. For a 200-SKU store it's a project; for a 15,000-SKU catalog it's a pipeline — generation, validation, and maintenance as the catalog changes. This is the layer where DIY stalls.

4. Distribute through AI merchant programs

ChatGPT's merchant program, Perplexity's, Google Merchant Center's AI Mode, Shopify's agentic plan, and the Universal Commerce Protocol underneath them. Each has its own feed requirements — all of them depend on layer 3 being done right.

5. Enrich and monitor

Normalized attributes, quotable titles and descriptions, FAQ blocks that models cite verbatim — then continuous monitoring of your share of answer so you know what's working.

The honest summary

Layers 1 and 2 are yours to do this week (our audit tells you exactly what's broken). Layers 3–5 at catalog scale are why our managed service exists: our technicians run the full program in 90 days, with a results guarantee. Either way, the first step is the same — find out your grade.