The 7-second window: why agentic commerce makes your PDP the new email subject line

For twenty years the job of a product page was to win the shopper's eye. The hero shot, the bullet list, the trust badges, the reviews — every component was designed for a human scanning a screen, weighing whether to click the green button.

That job has changed. Inside agentic commerce, the PDP is rarely the first surface a shopper sees and often is not a surface at all. The shopper has a conversation with an agent. The agent retrieves your PDP, reads it, and writes a paragraph that contains your brand or doesn't. The paragraph is what the shopper sees. The PDP is what the agent quotes from.

That difference is the most important shift in ecommerce UX since responsive design. It re-orders every component on the page. It changes which content matters, where it should sit, and how it should be written. This piece is about what the new PDP looks like — and the five-step rewrite we run with brands shipping into ChatGPT, Claude, Perplexity and Gemini today.

The token budget — your PDP is competing for ~700 words

Start with the math. When an agent receives a shopping query, it retrieves between four and twelve candidate pages, depending on the engine. Each candidate is summarised into a working memory the agent can quote from. The working memory budget per candidate is approximately 700 tokens — call it 500 English words.

That is the entire quote frame for your PDP. Five hundred words. Out of those, the agent will quote — verbatim or near-verbatim — somewhere between 30 and 90 words in the final answer to the shopper. The rest is context the agent uses to decide whether to mention you at all.

What this means in practice: every PDP component that doesn't fit into the first 500 retrievable words is wasted on the agent. Your scroll-and-zoom hero image carousel? Wasted. Your tabbed accordion of size, materials, care instructions, FAQs? Mostly wasted — only the first tab loads server-side on most platforms. Your "people also bought" rail? Wasted.

This is not theoretical. Run the experiment. Take a competitor's PDP, paste it into Claude or ChatGPT with the prompt "summarise this page in 100 words to recommend it to a friend". Watch which sentences it picks. Now do the same with your PDP. You will see the gap immediately.

What agents quote, and what they skip

Across about 1,200 prompts run against four engines over the last 90 days, we kept track of what made it into the answer for product-recommendation queries. The pattern is consistent enough to write rules from.

Table 1. Quote rate by PDP component type, across 1,200 product-recommendation prompts and four engines.

PDP component · Quote rate · Why it works (or doesn’t)

PDP component: Specific, factual descriptor sentences · Quote rate: 74% · Why it works (or doesn’t): Agents prefer claims they can re-state without paraphrase.

PDP component: Customer review excerpts (verified) · Quote rate: 61% · Why it works (or doesn’t): Authentic voice that the agent borrows directly.

PDP component: Material / spec lists · Quote rate: 58% · Why it works (or doesn’t): Quantifiable, comparable, structured.

PDP component: Use-case bullets ("works well for X") · Quote rate: 47% · Why it works (or doesn’t): Maps directly to shopper intent.

PDP component: Brand-voice marketing copy · Quote rate: 12% · Why it works (or doesn’t): Adjective-heavy, vague, low fact density.

PDP component: Hero image alt text alone · Quote rate: 4% · Why it works (or doesn’t): Agents prefer text written for humans, not crawlers.

The headline finding: factual specificity wins. The kind of marketing copy you wrote in 2020 — "discover effortless elegance with our signature collection" — is doing worse in agentic commerce than the boring spec sheet you cut from the page two redesigns ago.

The corollary: the brand voice is not dead, but it is no longer the first job. Your first job is to make the PDP citable. Brand voice gets the second pass, after the agent has decided to mention you.

The new shape of a PDP

Across the brands we work with, the PDPs that win in agentic commerce share a common structure. They are not all built the same way — Shopify themes vary, Wix templates vary — but the content order has converged. Here is the shape:

1. A single-sentence factual descriptor under the H1. "100% French linen unstructured blazer, oat colour, machine washable at 30°C." Not "effortless elegance". This sentence shows up in roughly 70% of agent quotes.
2. A four-bullet "works well for" block. Maps shopper intents (occasion, season, body type, comparison context) directly to the product. Each bullet should be a quotable sentence.
3. A material + care spec list, server-rendered. Six to ten rows max. Include units. "100% linen, OEKO-TEX 100 certified, machine washable 30°C, line dry."
4. Three to five verified review excerpts — server-rendered, with name + verified-buyer tag. The first one should match the most common product objection (sizing, wrinkling, fit).
5. A structured Q&A block, five questions. Authentic answers from your team. Schema.org/FAQPage emits cleanly.
6. A "comparison" paragraph. Two sentences positioning the product against the obvious alternative ("our blazer vs. wool: linen breathes better in summer, but wrinkles. Pick linen for travel; pick wool for office heating systems.").
7. A short shipping + returns paragraph in plain prose. Not a tab, not a popover.

Notice what is not in the top 7: the image carousel, the size-chart modal, the related-products rail. Those still matter for human shoppers and should stay on the page. They should also sit below the seven blocks above so the agent's first 500 words are spent on the seven blocks above instead.

Five rewrites, before and after

These are real rewrites from brands we work with, paraphrased lightly for privacy. The "before" version is the kind of copy a 2022 brand book would have approved. The "after" is what's running in production today.

Rewrite 1 — The opening sentence

BEFORE: "Crafted with care, our signature linen blazer brings effortless style to your wardrobe — perfect for any occasion."

AFTER: "Unstructured single-button blazer, 100% French linen, oat colour, fits true-to-size up to a 38" chest, machine washable at 30°C."
— PDP opening sentence rewrite. Agent quote rate: 14% → 68%.

The "before" sentence is a marketing string. It contains zero facts an agent can re-state. The "after" sentence is a paragraph any agent will quote verbatim. We lost the aspirational tone; we won the agent's working memory.

Rewrite 2 — The use-case bullets

BEFORE:
• Premium materials
• Timeless design
• Versatile fit
• Easy to style

AFTER:
• Travel: linen wrinkles, but recovers on a hanger overnight. Pack for 5-day trips.
• Hot weather: linen breathes 4× better than wool — comfortable above 24°C.
• Office: pair with a structured tee underneath for meetings; unbuttoned over a dress shirt for client lunches.
• Body shape: unstructured cut sits softly on shoulders; the half-belt back keeps the waist defined up to 92cm.
— Use-case bullets rewrite. Agent quote rate per bullet: 8–11% → 41–58%.

The "before" bullets are advertising. The "after" bullets are facts an agent can attach to a shopper's stated context. "I'm going on a 5-day trip to Greece" maps to bullet 1 + 2. "I need something for client meetings" maps to bullet 3. The mapping is the whole job.

Rewrite 3 — The review excerpt placement

Most PDPs paginate reviews behind a "show more" button that requires JavaScript. Half the crawlers that matter for agentic commerce do not execute JavaScript on retrieval. That means your reviews are invisible.

The fix is mechanically simple. Render the first 5 reviews server-side. Pick them by a rule, not by recency: one each addressing sizing, durability, fit, value, and the most common complaint. The agent reads them. It quotes from them. Citation rate moves within a week.

Rewrite 4 — The comparison paragraph

BEFORE: (no comparison content on the PDP)

AFTER:
"How this compares: vs. wool, linen breathes better in summer but wrinkles more. Pick linen for travel and warm-weather wear; pick wool for office buildings with strong heating. vs. our Hampton jacket: this one is unstructured and softer; Hampton is structured and reads more formal."
— Comparison paragraph addition. Agent quote rate: n/a → 47%.

Agents are often asked "X or Y, which should I buy?" The comparison paragraph turns your PDP into a candidate answer to that prompt. Without it, the agent goes to a third-party review site for the comparison and cites them, not you.

Rewrite 5 — The structured Q&A

The mistake teams make here is faking the Q&A — writing the questions and the answers in marketing voice and shipping them. Agents detect this. The signal is the absence of specifics. "How do I care for this product?" answered with "easy to care for, follow the label" reads as marketing. "Cold wash at 30°C, line dry, iron on linen setting if needed; avoid tumble dry — the heat sets wrinkles" reads as truth.

The fix is to mine your customer service inbox. Every brand has 50 real questions buyers actually ask. Answer five of them honestly on the PDP. That's it.

The PDP rewrite worksheet

To make this concrete, we built a worksheet teams can walk through in an hour per SKU. It's a five-step checklist with a scoring rubric, designed to be printed and filled in pen. By the end of the hour you have a rewrite plan you can hand to your copy team.

The worksheet is a Markdown file so it works in Notion, Google Docs, Linear, or just a text editor. Print it out for the hands-on hour; share the digital version with the team that ships the rewrites.

The two objections we hear

When we run this rewrite with brands, we get the same two objections, in this order.

Objection 1 — "This will hurt our brand voice." It does, slightly, on the opening sentence and the use-case bullets. The opening sentence becomes a fact. The bullets become utilitarian. The brand voice retreats one paragraph down the page, where the descriptive copy and the storytelling still live. What we have observed: on every brand that has run the rewrite for 60+ days, conversion rate on the page itself either stayed flat or went up. The brand voice cost was real; it just didn't matter at the conversion level.

Objection 2 — "Won't competitors copy our facts?" Yes, immediately. The facts are not the moat. The product is the moat. The PDP is just a faithful transcription of what the product is, optimised for the new reader. If a competitor copies your facts and lies about their product, the reviews will sort that out within a quarter.

Closing — the new attention economy

The reason this piece is called "the 7-second window" is that the agent reads your PDP in roughly seven seconds — fast for a retrieval, slow for an attention budget. In that seven seconds it either accumulates enough quotable, factual content to mention you in the answer, or it moves on. There is no rich design that compensates. There is no animation that helps.

The brands that win the next two years of ecommerce are the ones whose PDPs accept this. They drop the marketing copy from the top. They write factual descriptors. They server-render reviews. They put comparison paragraphs on the page. They publish honest Q&A.

None of this is glamorous. All of it is necessary. Print the worksheet. Pick three PDPs. Spend Tuesday morning rewriting them. See what happens to citation rate by Friday.