How to write PDP copy for AI search.

Why most PDPs lose at AI search

Most Shopify PDPs were written for two audiences: the person browsing the storefront and Google's old keyword-matching algorithm. Neither of those audiences is dominant on AI search. ChatGPT reads the page to extract a quotable passage that answers a specific buyer question. Perplexity reads it to corroborate a claim against three to seven other sources, per eCommerce Insights's manual review of 200 shopping queries in Q1 2026. Both engines reward density of factual passages and punish brand-voice softness. Most PDPs do the opposite: they open with poetry, bury the specs in tab six, and treat the FAQ as an afterthought.

The 300-character rule

The first 300 characters of the PDP body have outsized weight in AI retrieval. They are the passage most likely to be summarized into an answer. They are also the passage a brand is most tempted to fill with brand-voice opening lines. Resist. Use that space to answer the single most common buyer question with three quotable facts.

A brand-voice opener: "The Wanderer is a beautifully crafted weekend bag designed for the modern traveler who demands quality and style."

A fact-density opener: "The Wanderer is a 35L waterproof weekend bag for two-night trips. 1000-denier nylon shell, YKK zippers, rated to IPX5, certified carry-on for Delta and American. Lifetime repair guarantee."

The second opener can be cited verbatim. The first cannot. That is the entire difference.

Claims become facts

The conversion is mechanical. Walk through the PDP and replace every adjective with a number, a material, a certification, or a duration. "Durable" becomes "1000-denier ballistic nylon." "Long battery life" becomes "47 hours per charge." "Comfortable" becomes "memory-foam insole, tested to 30 miles in a 2025 Backpacker field review." Every conversion adds one citeable passage to the page.

Headings as buyer questions

An H3 that reads "Materials & Construction" describes the section to a brand manager. An H3 that reads "Is this fully waterproof?" describes the section to a buyer and to the AI engine reading the page. Engines map heading text to likely user queries, so headings phrased as questions strengthen the page's match to those queries. Use H3s for the top three to five questions a buyer types into ChatGPT for your category. The eCommerce Insights AEO Grader evaluates this directly under the heading-hierarchy score.

Third-party grounding

A PDP that says "rated waterproof" is weaker than one that says "rated IPX5, verified in 412 customer reviews and a 2025 Outside Magazine field test." Visible review counts, syndicated review-site coverage, structured Review schema, and external publication mentions all add grounding signal. ChatGPT and Perplexity both penalize claims that have no independent corroboration; Perplexity particularly so. See how to test whether Perplexity recommends your product for the measurement angle.

A before-and-after rewrite

Illustrative — a real D2C outdoor brand's flagship parka, simplified.

Before: "Embrace the elements in the Northwest Parka, our signature heavyweight outerwear piece designed for the discerning urban explorer. Crafted with the finest materials and an obsessive attention to detail, this is more than a coat — it's a statement of intent."

After: "The Northwest Parka is a 650-fill-power down jacket for Pacific Northwest winters. Waterproof to 10,000 mm, windproof, with a YKK two-way zipper, adjustable storm hood, and seven external pockets. Rated for sustained temperatures down to -20°F in a 2025 Outdoor Gear Lab test. Lifetime repair warranty."

The After version answers four buyer questions in 300 characters: what is it, how warm, how waterproof, how trusted. The Before version answers none. Run both through the AEO Grader and the score differential is consistently 20 to 35 points.

"Durable" is a claim. "1000-denier ballistic nylon, tested to 30,000 cycles" is three facts and a citation.

The eight-item rewrite checklist

1. First 300 characters answer the top buyer question with three quotable facts.
2. Every adjective replaced with a number, material, certification, or duration.
3. Top three to five H3 headings phrased as buyer questions, verbatim where possible.
4. Each H3 followed by a 40-to-80-word answer with at least one quotable fact.
5. Visible review count, star rating, or third-party endorsement on the page.
6. FAQPage block with five real Q&A; questions match buyer language.
7. Product JSON-LD in head with brand, sku, gtin (if available), aggregateRating.
8. No duplicate boilerplate; shipping and warranty language linked, not repeated.

A PDP that ships all eight scores above 75 on the AEO Grader's five-dimension rubric in eCommerce Insights's experience with mid-market D2C catalogs through Q1 2026.

Frequently asked questions

Related jobs

Related tools

• AEO Grader — score the rewrite before and after.
• Shopify SKU Visibility Grader — measure citation lift.