MarketingEcommerceIntermediate20 minSaves About 45–60 minutes on a prepared SKU; review time varies

Product Description Optimisation Audit

Turn an existing product description into quoted, prioritised edits and a publishable draft that preserves your voice.

Audit product copy against verified specifications, buyer needs, search intent and competitor context. Get exact replacements, unsupported-claim flags and a rewrite applying only high-impact changes.

Ready-to-use prompt

The prompt

Copy it as-is, then swap the bracketed placeholders for your own details before running it.

prompt.txt
Role: You are an ecommerce conversion editor auditing a product description.

Context:
- Current description: {{current_description}}
- Product specifications: {{specs}}
- Target buyer: {{buyer}}
- Target keyword: {{keyword}}
- Competing product pages: {{competitors}}
- Known objections and return reasons: {{objections}}

Task: Return an audit table with:
1. Issue
2. Current text, quoted exactly
3. Suggested replacement, written out in full
4. Why the replacement is better (clarity, evidence, objection, scannability, search)
5. Expected impact (high / medium / low)
6. Effort (S / M / L)

Then return:
- The rewritten description in full, applying only the high-impact fixes
- Anything in the current copy that is working and must be preserved
- Any claim in the current copy that the specs do not support

Rules:
- Every row must quote real current text, no generic advice such as "add more detail".
- Do not change the brand voice, match it.
- Do not add a keyword where it does not read naturally.
- Never remove a legally required statement.

Finish with: the three changes to make first if only thirty minutes are available.

Estimated results

DifficultyIntermediate
Setup time20 min
Time savedAbout 45–60 minutes on a prepared SKU; review time varies
Best modelsChatGPT, Claude, Gemini
Best audienceEcommerce, Retail

Editor's note

Why this prompt matters

A useful product-description audit should leave an ecommerce manager with decisions, not a list of writing principles. This prompt anchors every recommendation to an exact quotation, then supplies replacement copy that a reviewer can approve or reject.

Start with one SKU and paste the complete description into {{current_description}}. Supply catalogue facts in {{specs}}, including dimensions, materials, compatibility, care instructions and applicable warnings. These facts establish what the rewrite can substantiate. Describe the buying situation in {{buyer}}: a commuter choosing a leak-resistant mug needs different information from someone buying a desk accessory.

Use {{objections}} for recurring support questions and return reasons, not imagined anxieties. Include required wording and any marketplace formatting restrictions within the supplied context; there is no separate platform variable. Without those constraints, a readable rewrite may still be unsuitable for publication.

Anatomy

Prompt engineering breakdown

Role

The conversion-editor role frames the task as evidence-led revision, not unrestricted promotional writing.

Context

Specifications establish factual boundaries; buyer, keyword, competitors and objections establish relevance without authorising borrowed claims.

Goal

Produce implementable replacements and a restrained full draft, not a general copywriting critique.

Constraints

Exact quotations, matched voice, natural keywords and protected legal statements constrain unnecessary or risky changes.

Output format

The table supports approval; the rewrite supports implementation; preservation and claim lists support quality control.

Why this structure works

Diagnosis comes before rewriting, making the final draft easier to compare against approved priorities.

What you'll get

Expected output

Expect a six-column audit table, followed by the complete description with only high-impact fixes applied, a preservation list and unsupported claims. The response should finish with three changes for a thirty-minute editing window.

For example, suppose the current mug description says “Made for busy mornings.” and verified specifications confirm a 450 ml capacity, stainless steel construction and hand-wash-only care. A useful row could be:

| Issue | Current text | Suggested replacement | Why better | Impact | Effort | |---|---|---|---|---|---| | Basic selection facts missing | Made for busy mornings. | Made for busy mornings, this 450 ml stainless steel travel mug is hand-wash only. | Adds capacity, material and a care limitation buyers need before ordering. | High | S |

If the same description promises “Keeps drinks hot all day” without supporting thermal-test evidence, that belongs in the unsupported-claims list. Do not replace it with another invented duration. The impact rating is a hypothesis, not measured uplift; the example's wording still needs a brand-voice check.

Under the hood

Why this prompt works

The quotation requirement makes the audit traceable: reviewers can find each passage and compare the replacement directly. Full replacements expose whether the model has actually solved the issue rather than merely labelled it.

Separating impact from effort supports practical prioritisation. Define S as a direct copy edit, M as an edit needing factual confirmation, and L as work requiring wider approval or page changes. The high-impact-only rewrite limits unnecessary churn, while the preservation list protects useful specificity and established tone. Neither mechanism proves that a sentence converts; analytics and customer evidence remain necessary.

Model fit

Best AI models for this prompt

Claude

Use Claude as the primary drafting option for a voice-sensitive audit. Provide the full description rather than asking it to infer tone from a brand name. Check that the final rewrite excludes medium- and low-impact edits.

ChatGPT

Useful for iterative review of individual rows. Keep the six-column format explicit and verify every quotation against the source; plausible paraphrases do not satisfy this prompt.

Gemini

Useful when comparing supplied competitor excerpts alongside product facts. Label each source clearly so competitor specifications do not migrate into your rewrite. Across all three models, browsing availability varies: a pasted URL does not establish that its contents were accessed.

When to use

  • Before sending campaign traffic to a priority SKU with vague supplier copy.
  • When support tickets repeatedly reveal confusion about sizing, compatibility or care.
  • When a page attracts relevant visits but shoppers appear to lack decision-critical information; inspect other conversion barriers too.

When not to use

  • For an empty description: there is nothing to quote and replace.
  • For unresolved safety or regulated claims without qualified review.
  • As a substitute for fixing stock availability, delivery costs, broken variants or weak imagery.
  • For unattended catalogue-wide publishing.

Get more from it

Pro tips

  • 1

    Label competitor excerpts with their source and date; use them to understand buyer expectations, never as evidence for your product.

  • 2

    Supply one primary keyword and its buying intent. Reject insertions that make an otherwise clear sentence awkward.

  • 3

    Check unsupported claims before prioritising conversion edits. Escalate conflicts involving legally required wording rather than letting the model silently remove it.

  • 4

    Record the publication date and changed passages. Compare relevant funnel metrics over sufficient traffic; two weeks alone does not establish a reliable result.

Don't ship this

Common mistakes

  • ✗ Accepting invented quotations.

    Fix — Search the source description for each quoted passage before approving its replacement.

  • ✗ Treating impact labels as forecasts.

    Fix — Prioritise using customer evidence and page data, then validate through testing where feasible.

  • ✗ Supplying only competitor URLs.

    Fix — Paste relevant excerpts unless page access has been confirmed.

  • ✗ Publishing every suggested edit.

    Fix — Keep the rewrite limited to high-impact fixes and review all factual changes.

People also ask

Frequently asked questions

Q.Will it rewrite my whole product page?

Only if you ask it to. The default output is a table of targeted fixes that quote your current text, plus a rewrite applying just the high-impact ones, so you keep whatever is already working.

Q.How many changes should I ship at once?

Three. Applying everything at once means you cannot tell which change moved the metric, and on a low-traffic store you will never separate the effects afterwards.

Q.What if the audit says my copy claims something unsupported?

Fix it before anything else. Unsupported claims drive returns, disputes and, in some categories, regulatory attention — this is the one section of the audit that is not optional.

Version 1.1Last reviewed September 21, 2026
Reviewed by PromptInFlow Editorial Team