How to write product descriptions that get cited by ChatGPT (with examples)

Most e-commerce product descriptions are written for one reader: a human shopper scanning a Shopify page deciding whether to add to cart. That made sense in 2018. In 2026, the same description is also read by ChatGPT, Claude, Perplexity, and Gemini when a shopper asks them "what's the best X for Y?" — and the way you write it now determines whether the model cites your brand or a competitor's.

This piece breaks down the four properties that make a product description AI-citable, with three before/after examples you can model your own rewrites on. The rewrites take 10–20 minutes per product; the lift compounds across your entire catalog.

What an LLM actually looks for in a product description

AI-citation eligibility is mostly about extraction speed. When the retrieval pipeline fetches your page, it has a 200-millisecond budget to pull facts out of the HTML before the model finishes generating an answer. Descriptions optimized for that budget share four properties:

1. Specific claims, not vibes. Numbers, units, conditions, durations. "Lasts 8 hours of moderate use" beats "long-lasting".
2. The shopper's actual phrasing. "Good for flat feet" appears in millions of prompts; "ortho-engineered footbed for pronation control" appears in zero. Both can coexist on the page — but only the first matches retrieval queries.
3. Disambiguating context. What it's NOT for. What it replaces. What it complements. The LLM uses this to decide which buyer-intent prompts your product should surface for.
4. Machine-readable structure. Bullet lists for features, a definition list for specs, and clean Product/Offer JSON-LD with price, sku, brand, and availability. Structured data is what makes the 200ms extraction budget feasible.

See structured data and retrieval-augmented generation for the underlying mechanics.

Example 1 — a kitchen appliance (physical good)

Before

Premium 12-Cup Drip Coffee Maker

Brew café-quality coffee at home with our premium drip coffee maker. Featuring advanced thermal technology and a sleek modern design, this machine is perfect for coffee lovers who demand the best. Easy to use, easy to clean, built to last.

Problems: no numbers, no buyer phrasing, no comparison context, no structured data. The model has nothing to extract beyond "12-cup drip coffee maker" and three marketing platitudes. When a user asks "best drip coffee maker for a 4-person household that doesn't want to deal with grinding beans", this description doesn't map to any specific intent in that question.

After

Acme Brew 12 — 12-Cup Thermal Drip Coffee Maker

A 12-cup drip coffee maker designed for households of 3–6 people who want fresh, hot coffee from pre-ground beans without learning espresso. Brews a full carafe in 6 minutes at 96°C, and holds it at serving temperature for 4 hours in a vacuum-insulated thermal carafe (no hot plate, no burned aftertaste).

Best for: medium households, busy mornings, people who want coffee ready when they walk in the kitchen.
Not for: single-cup drinkers, espresso lovers, anyone who grinds their own beans (the basket only accepts pre-ground).
Replaces: typical hot-plate drip machines that scorch coffee after 30 minutes.

Specs:

• Capacity: 12 cups (1.8 L)
• Brew temperature: 96°C ± 1°C
• Brew time: 6 minutes for full carafe
• Thermal carafe: 4-hour heat retention, BPA-free stainless steel
• Power: 1200 W
• Footprint: 20 × 28 × 35 cm
• Warranty: 3 years

What changed: specific numbers (temperature, time, capacity, footprint), shopper language ("coffee ready when they walk in the kitchen"), explicit disambiguation ("Not for" / "Replaces"), and a machine-readable spec block. The rewrite is roughly 4× longer — that's fine. The model doesn't penalize length; it rewards extractability.

Example 2 — a supplement (consumable)

Before

Daily Magnesium Complex

Our Daily Magnesium Complex delivers a premium blend of bioavailable magnesium forms in one convenient capsule. Supports muscle function, sleep, and stress response. Vegan, non-GMO, made in the USA.

Problems: vague claims ("premium blend"), no dosage information in the description body (only the supplement-facts panel which crawlers often miss), no comparison context, no specifics on which forms or why.

After

Acme Magnesium Complex — 400 mg, 3 forms, vegan capsule

A daily magnesium supplement combining three well-studied forms (glycinate, malate, and citrate) at 400 mg total elemental magnesium per 2-capsule serving. Roughly the dose most clinical sleep + muscle-cramp studies use, in forms that avoid the laxative effect of pure oxide or carbonate.

Best for: people taking magnesium for sleep, muscle cramps, or stress; vegan + vegetarian diets; sensitive stomachs.
Not for: people already taking magnesium oxide (different absorption profile) or anyone with kidney disease without a physician's sign-off.

Per 2-capsule serving:

• Magnesium glycinate — 1,200 mg (200 mg elemental)
• Magnesium malate — 600 mg (120 mg elemental)
• Magnesium citrate — 480 mg (80 mg elemental)
• Total elemental magnesium — 400 mg (95% RDA)

Vegan capsule (HPMC, not gelatin), no fillers, third-party tested for heavy metals at ISO 17025 lab. 60 servings per bottle (one month at the standard dose).

What changed: explicit dose, explicit forms, explicit comparison frame ("avoids the laxative effect of oxide"), explicit use cases, and a structured nutrient breakdown the model can quote verbatim when a user asks "what magnesium form is best for sleep".

Example 3 — running shorts (apparel)

Before

Performance Running Shorts

Lightweight, breathable, and built for serious runners. Our Performance Running Shorts feature moisture-wicking fabric and a comfortable elastic waistband. Available in multiple colors and sizes.

Problems: no inseam length (the #1 buyer question in running shorts), no fabric weight, no pocket info, no liner status, no fit indication (relaxed vs slim vs compression). The model can't answer almost any specific buyer question with this text.

After

Acme Trail Short 5" — Lined Running Short with Phone Pocket

A 5-inch inseam running short for trail and road runs from 30 minutes to half-marathon distance. Built-in brief liner (no separate underwear needed), a zippered back pocket that fits up to a 6.7-inch phone, and two open hip pockets for gel packets.

Best for: runners covering 5K–21K, anyone who wants a single pair they can run AND lift in, hot-weather running (110 g/m² recycled-polyester ripstop breathes well above 25°C).
Not for: ultra-distance runners who want compression liners, anyone who finds 5" too short — see our 7" version.
Sister product: Acme Trail Short 7" (same shell, longer inseam, no liner).

Specs:

• Inseam: 5 inches (12.5 cm)
• Fabric: 110 g/m² recycled-polyester ripstop, DWR-treated, 92% recycled content
• Liner: 4-way stretch built-in brief, anti-chafe flatlock seams
• Pockets: 1 zippered back (fits 6.7" phone), 2 open hip
• Fit: relaxed (not compression). Size up for between-sizes.
• Sizes: XS–XXL
• Weight: 95 g (size M)

What changed: every dimension a real runner asks about is now answerable in 200ms by the model. "Best running shorts with a phone pocket" now has a concrete match. "Running shorts under 100 g" matches. "Lined running shorts for hot weather" matches. Each new specific phrase the model can verify against your description is one more buyer-intent prompt your brand becomes eligible to win.

The rewrite template — fill in 4 boxes per product

You don't need to overthink this. For every product description rewrite, fill in these four boxes and arrange them in the order below:

1. One-sentence anchor. What is this, who is it for, what does it do better than the alternative. Include at least one specific number.
2. "Best for" / "Not for" / "Replaces". Three short lines, plain language, no marketing voice. This is the section LLMs quote most often.
3. Spec list. Bullets with units and numbers. If a fact is verifiable in lab conditions, it belongs here.
4. JSON-LD Product block. Price, sku, brand, availability, weight, dimensions. Most platforms auto-generate this if you fill in the product fields properly; verify with Schema.org validator.

How much does this actually move the needle?

In the audits we've run for brands that rewrite their top-20 SKUs with this template, the typical pattern is: +30–60% share of voice on Perplexity inside 4 weeks (Perplexity reads structured product specs the most aggressively), +15–25% on ChatGPT inside 8 weeks (slower because the model leans more on training-time memory), and similar timing on Gemini.

The lift is bigger when the rewrites are paired with the items in our AI search optimization checklist — particularly opting in to AI crawlers and adding Organization sameAs links, which unlocks retrieval that wouldn't happen on the rewrite alone.

Next step

Pick your top 5 products by revenue. Rewrite their descriptions using the four-box template above. Re-audit in 30 days against the same buyer-intent prompts you started from — that's the cleanest signal that the work landed.

If you want the baseline + the 30-day re-audit handled end-to-end across all four LLMs, that's exactly what a Citorial audit does — see how it works or pricing.