The D2C playbook for AI search, 2026 edition.

AI-search practices are still consolidating. That's the first thing to say before presenting a seven-step plan. The second is that some things are clear enough to act on — schema completeness, answer-coverage rewrites, per-SKU monitoring — and it's worth acting on them now rather than waiting for industry consensus. The playbook below is what eCommerce Insights recommends Shopify D2C brands do this quarter. Each step includes a time estimate and a caveat where eCommerce Insights is less certain.

Step 1 — Audit your catalog's AI visibility per-SKU

The wrong first move is rewriting PDPs. The right first move is data collection. Before touching any page, run each of your top-revenue SKUs through a fixed prompt set on ChatGPT, Perplexity, and Google AI Overviews. Note which SKUs are named, which are silent, and which are mischaracterized. Manual audit works up to roughly 30 products; beyond that, automation (eCommerce Insights or otherwise) becomes the realistic option. The output is a prioritized list of SKUs by visibility gap weighted by revenue.

See the blog post on auditing ChatGPT visibility manually for the prompt framework. Expect this step to take one to three days for a focused team.

Step 2 — Fix your llms.txt

Serve a llms.txt file at your domain root. eCommerce Insights's free generator reads your Shopify catalog and outputs a compliant version; upload it, verify the path is reachable. This step is cheap and mostly done in an afternoon. Its effect is small but real — some AI crawlers respect the file as of Q1 2026, and the cost to serve it is near zero. Treat it as hygiene rather than leverage.

Step 3 — Fill your Product JSON-LD blank fields

Run your top-revenue SKUs through Google's Rich Results Test. Note which fields are missing. For Shopify, the most common gaps are GTIN, brand, material, country_of_origin, and aggregateRating (only include aggregateRating if real). Add each as a metafield if it doesn't exist, populate, wire into your theme's JSON-LD template. This step rewards effort proportional to catalog size — allocate a week for a mid-size catalog, two for a larger one. See the companion post on Shopify metafields.

Step 4 — Rewrite your thinnest 20 percent of PDPs

Every brand has PDPs where the description is under 120 words and the page exists mostly to host a product image. Those are the fastest to fix. For each, add three blocks: a use-case paragraph that describes who the product is for and what it solves, a comparison note against a known alternative (honest about where you're smaller or larger), and three buyer-phrased Q&A entries in FAQ schema. The goal is to make each PDP answer the question a buyer would ask, not just describe features.

Step 5 — Seed review sources with accurate product data

Because review sites dominate citations on most shopping queries, getting covered is an outsized lever. Identify the three to five publications most likely to cover your category — Wirecutter, The Strategist, category-specialist blogs. Build a press kit with clean GTIN, material, dimensions, and a high-resolution image. Pitch their category editor. Follow up on update cycles (many review roundups get refreshed annually). This is PR work, not software work. It's slow and often frustrating. It's also the step that compounds.

Step 6 — Monitor weekly, not quarterly

AI-engine behavior changes often enough that quarterly monitoring is too coarse. Weekly catches drift before it becomes a revenue problem. For a catalog of any meaningful size, weekly monitoring means automation; manual prompt runs don't scale. eCommerce Insights automates the cadence; other tools cover parts of the problem (brand-level only, or single-engine). Whatever you use, weekly is the cadence goal.

Step 7 — Triage by revenue, not by score

Once data flows, the temptation is to tackle the worst-scored SKUs first. Resist it. The right priority is the SKU with the largest gap between current visibility and revenue weight. A low-score SKU generating 15 percent of your D2C revenue matters enormously. A low-score SKU generating 0.5 percent of revenue is a rounding error. Your visibility-score dashboard should sort by revenue-weighted gap, not by raw score.

This is the sequencing insight most teams get wrong in their first quarter. eCommerce Insights surfaces revenue weight alongside score so triage lands on the right SKU.

What this playbook won't do

It won't make you the most-cited brand in your category overnight. It won't compensate for a product that genuinely doesn't compete on its own merits. It won't protect you from an AI engine's retrieval stack refresh that shifts the rules mid-quarter. Those are limits worth naming. What it will do is improve your chances that the right SKUs surface on the queries your buyers actually run — steadily, over months, as a compounding practice.

Key takeaways

• Start with an audit of top-revenue SKUs, not a catalog-wide rewrite.
• llms.txt is hygiene; do it, then move on.
• Fill Product JSON-LD blanks — GTIN, material, brand, country_of_origin — as a catalog-wide sweep.
• Rewrite the thinnest 20 percent of PDPs to answer buyer questions, not just describe features.
• Pitch review sites; their coverage amplifies your citations on most shopping queries.
• Monitor weekly — the cadence most teams skip.
• Triage by revenue-weighted gap, not raw visibility score.

Related reading