AI commerce
Put product identity before brand story in the opening description
Use the first part of a product description to explain fit and key facts, while treating the reported 40-word shopping cutoff with proper limits.
TL;DR
- Put the product type, intended use and decisive specification near the start of the description.
- Peec reported a 50-token cutoff in a specific ChatGPT shopping observation; this is not a universal rule for all AI systems or every page.
- Use a short opening as an editorial stress test, while keeping the full description and structured product fields accurate.
- Check the actual feed output as well as the website: they may contain different text.
A good product description tells buyers what the item is before it tells the brand story. This helps when the text appears in a short preview, a feed or a list of options. Can someone tell what is for sale and whether it fits their need?
A fixed word count is a less useful goal. Names, units and language affect how a model splits text into tokens. A short opening can be clear without implying that a model ignores the rest of the page.
What does the reported cutoff actually establish?
In an October 16, 2026 article, Peec described debug-data observations of ChatGPT shopping. It reported that descriptions in a product prompt block were cut at 50 tokens, roughly 40 English words. The study discussed different US and European retrieval paths. It did not publish a universal OpenAI rule for all content ingestion. Treat it as a dated observation of that shopping setup. See Peec's original analysis.
Google Merchant Center has its own description specification. Its guidance calls for product features and technical detail and says to place important information first. That is a separate product-feed context, not independent proof of the ChatGPT cutoff. Check the current Merchant Center description requirements.
The practical overlap is useful: do not bury identity and fit beneath generic praise. The platform-specific limits still need separate verification.
Worked example: turn an opening into a usable product answer
This example is fictional. A seller offers the Dock D6, a USB-C dock. The approved specification sheet says it works with laptops that support DisplayPort Alt Mode, provides HDMI output and supplies up to 85 W to a compatible laptop. An existing description begins with the company's history and a promise of “limitless productivity.”
Check the spec sheet, box contents and exported feed. “USB-C” alone could make a buyer think the dock works with a laptop it does not support. Keep the port requirement near the start, even if that leaves less room for a sales pitch.
A revised opening is:
Dock D6 is a USB-C desk dock for laptops that support DisplayPort Alt Mode. It adds HDMI, wired Ethernet and USB connections, with up to 85 W laptop charging when used with the required power supply.
This is a writing sample, not a real product claim. The full page must still explain supported display modes, included accessories, power-supply requirements and tested limitations.
| Buyer question | Opening answer | Required check |
|---|---|---|
| What is it? | A USB-C desk dock. | Category agrees with the product record. |
| Will it fit my setup? | The laptop must support DisplayPort Alt Mode. | Technical owner confirms the compatibility condition. |
| What does it add? | HDMI, wired Ethernet and USB connections. | Ports match the exact variant and images. |
| What qualifies the benefit? | Charging needs the required power supply. | Full copy states whether that supply is included. |
| Is the feed consistent? | The exported text starts with these same facts. | Inspect the outgoing record, not only the CMS preview. |
The product owner checks the facts. The editor checks the wording. The feed owner checks what is sent. The opening passes if a reader can name the item and its main limit. It fails if a key caveat is cut just to meet a word count.
Keep the opening concise and the record complete
Use plain nouns and units. Replace “best-in-class versatility” with the actual supported use. Compare a rival product near the start only when the claim is useful, backed by evidence and allowed in that feed. Do not invent a claim that the product is better just to fill a formula.
Keep price, stock, identifiers and variants in the appropriate fields. A short description is not a reason to flatten the full catalog into one paragraph. Different surfaces may use different fields, and an export mapping can silently replace carefully edited website copy.
Start with products that attract fit questions or costly returns. Review a manageable sample, note recurring failures and then change the shared template. Count fixed records as work done. Track useful visits or sales as a separate result.
Common questions
Should the description stop after 40 words?
No. Use the opening test to find buried facts. Keep the detail a buyer needs, and follow the actual destination's current specification.
Do tokens equal words?
No. Tokens are model-processing units and may be whole words, word fragments or punctuation. Do not convert a reported token limit into an exact word rule for every language.
Semanta helps turn product knowledge into clear buyer-facing content and audit the evidence behind claims. If your catalog reads well in the CMS but becomes vague in feeds, discuss a product-content review with a sample page and export.