E-commerce · Conversion
The Best Prompt for a Product Page Audit
The best prompt for a product page audit reviews the page the way a shopper reads it. This prompt checks your pasted page content against five buying questions, from what the product is to how to buy it, and logs each problem with quoted evidence, a severity rating, a concrete fix and the effort involved. You get a prioritized top five and rewritten copy, plus a clear list of things the model could not assess, such as images and speed.
Product Page Audit Prompt
Audits pasted product page content against the five questions a shopper asks before buying, with quoted evidence, severity, fixes and rewrites.
- Works with
- ChatGPT, Claude, Gemini
- Difficulty
- Intermediate
- Complexity
- Multi-part reply
- Variables
- 5
- Version
- 1.0
Customize this prompt
Fill in what you know. Anything left blank stays as a [bracketed placeholder]. Nothing you type leaves your browser.
5 variables · fields marked * matter most
Act as a conversion-focused e-commerce consultant. Audit the product page below the way a first-time visitor experiences it: as a series of questions that must be answered before they will buy. # Context Product category: [Product category] Typical shopper and how they arrive: [Typical shopper and how they arrive] Store platform: [Store platform] Data or complaints I already have about this page: [Data or complaints you already have] # Page content <page> [Page content] </page> # Audit method Go through the five shopper questions in order. For each, judge only what the pasted content shows. 1. What is this? Title, opening lines, main image description. 2. Is it right for me? Specs, sizing, compatibility, use cases, variants, comparison with similar items. 3. Can I trust it? Reviews, guarantees, brand and seller information, clarity and honesty of claims. 4. What will it really cost and when will I get it? Price, shipping cost and time, returns, stock status, any fees. 5. How do I buy? Call-to-action wording, option selection, what happens after the click. For every problem found, record: - Finding: what is wrong or missing - Evidence: the exact text from the page, quoted, or "absent" - Severity: Blocker (stops a purchase), Doubt (creates hesitation) or Polish - Fix: a specific change, including replacement wording where copy is the problem - Effort: Low, Medium or High # Boundaries - Assess only the pasted content. You cannot see layout, images, speed or mobile rendering, so do not comment on them. List them under "Not assessed" with what I should check myself. - Do not estimate conversion-rate gains or cite industry benchmarks. Describe each fix as a hypothesis to test. - If I supplied data or complaints, connect findings to them where they plausibly relate, and say when the link is a guess. - Do not recommend adding reviews, badges, urgency timers or claims the store cannot honestly support. - Note what the page already does well in no more than three bullets. Do not pad with praise. # Output format 1. One-paragraph verdict: the single biggest reason a shopper would leave this page. 2. Findings table grouped by the five questions, with the columns above. 3. Top five fixes in priority order, weighing severity against effort. 4. Rewrites: an improved title, opening paragraph and call-to-action label, using only facts present on the page. 5. Not assessed: the checks that need a person looking at the live page.
Use it when you are
- Reviewing a product page that gets traffic but few add-to-carts
- Checking a new page before sending paid traffic to it
- Creating a prioritized fix list for a developer, copywriter or store owner
- Auditing a best seller to produce a template for the rest of the catalog
Why this prompt works
Five questions instead of a checklist of best practices
Generic audits recite best practices whether or not they apply. Organizing the review around what a shopper needs to know, in the order they need it, ties every finding to a reason someone might leave. It also makes omissions visible, which matter as much as errors.
Findings must quote the page
Requiring the exact text, or the word "absent", as evidence keeps the model from describing a typical product page instead of yours. If a finding quotes something that is not on your page, you know immediately to disregard it. The quotes also make the audit easy to hand to whoever edits the page.
Severity and effort are rated separately
A long list of equal-looking issues is hard to act on. Rating each finding as Blocker, Doubt or Polish, with effort alongside, lets the top five weigh impact against cost, so a missing returns policy outranks a wording tweak. Low-effort fixes for Blockers rise to the top, where they belong.
The model is told what it cannot see
From pasted text a model has no view of image quality, layout, load time or mobile behavior, yet it will comment on them if asked loosely. The boundaries move those items to a "Not assessed" list, giving you a manual checklist instead of guesses.
Fixes are framed as hypotheses
Nobody can know from reading a page how much a change will lift sales. Barring predicted gains and benchmarks, and ruling out manufactured urgency or unsupported trust badges, keeps recommendations to changes you can make honestly and then measure. Your own before-and-after numbers are the only reliable verdict on a fix.
How to use it
- Open the live product page, select all the visible text from top to bottom and paste it into the page content field. Add a short note for each image and video.
- Include shipping, returns and review text even if it sits in tabs or accordions, and say that it is hidden by default.
- Add any numbers or recurring customer questions you have for this page.
- Run the prompt and read the one-paragraph verdict, then the top five fixes.
- Verify each quoted piece of evidence against the page, and work through the "Not assessed" list yourself on a phone.
- Make one or two changes at a time and compare results over a sensible period before making more.
AI model compatibility
| Model | Status | Notes |
|---|---|---|
| ChatGPT | Expected to work | If your tool accepts images, you can add a screenshot and ask for layout observations as a separate follow-up. Keep the text audit first so findings stay tied to quoted evidence. |
| Claude | Expected to work | The page content sits inside <page> tags. Add a second block such as <reviews> if you want customer reviews considered separately from the store's own copy. |
| Gemini | Expected to work | Paste the page text in the order it appears on screen, top to bottom, so the audit reflects what a shopper sees first. |
“Expected to work” means the prompt uses plain, model-independent structure and has been reviewed for that — not that it has been run on that model. A model is marked Tested only when a test record is published below.
Worked example
Scenario: An illustrative furniture store reviewing a standing desk page that underperforms its other products.
Act as a conversion-focused e-commerce consultant. Audit the product page below the way a first-time visitor experiences it: as a series of questions that must be answered before they will buy. # Context Product category: Standing desks Typical shopper and how they arrive: Remote workers comparing three or four desks, mostly arriving from search ads on a phone Store platform: Shopify Data or complaints I already have about this page: Add-to-cart rate is lower than our other desks; support often gets asked about assembly time and whether the desk fits under a sloped ceiling # Page content <page> Title: ErgoRise Desk. Price: $549. Options: Top color (Oak, Walnut, White). Description: "Elevate your workday with the ErgoRise, engineered for peak productivity." Specs: Dual motor; 4 memory presets. Images: desk in a studio, close-up of control panel. Button: "Add to cart". Shipping: "Calculated at checkout". Returns: link in footer only. Reviews: star rating shown, no review text on page. </page> # Audit method Go through the five shopper questions in order. For each, judge only what the pasted content shows. 1. What is this? Title, opening lines, main image description. 2. Is it right for me? Specs, sizing, compatibility, use cases, variants, comparison with similar items. 3. Can I trust it? Reviews, guarantees, brand and seller information, clarity and honesty of claims. 4. What will it really cost and when will I get it? Price, shipping cost and time, returns, stock status, any fees. 5. How do I buy? Call-to-action wording, option selection, what happens after the click. For every problem found, record: - Finding: what is wrong or missing - Evidence: the exact text from the page, quoted, or "absent" - Severity: Blocker (stops a purchase), Doubt (creates hesitation) or Polish - Fix: a specific change, including replacement wording where copy is the problem - Effort: Low, Medium or High # Boundaries - Assess only the pasted content. You cannot see layout, images, speed or mobile rendering, so do not comment on them. List them under "Not assessed" with what I should check myself. - Do not estimate conversion-rate gains or cite industry benchmarks. Describe each fix as a hypothesis to test. - If I supplied data or complaints, connect findings to them where they plausibly relate, and say when the link is a guess. - Do not recommend adding reviews, badges, urgency timers or claims the store cannot honestly support. - Note what the page already does well in no more than three bullets. Do not pad with praise. # Output format 1. One-paragraph verdict: the single biggest reason a shopper would leave this page. 2. Findings table grouped by the five questions, with the columns above. 3. Top five fixes in priority order, weighing severity against effort. 4. Rewrites: an improved title, opening paragraph and call-to-action label, using only facts present on the page. 5. Not assessed: the checks that need a person looking at the live page.
What a good response should contain
- A verdict naming the main gap, most likely that the page never states the desk's dimensions, height range or weight capacity
- Findings that quote lines such as "Elevate your workday" and "Calculated at checkout" as evidence, each with severity and effort
- A link drawn between the missing height range and assembly details and the support questions supplied, labeled as a likely connection
- Rewrites that use only facts on the page and flag the specifications still needed instead of inventing them
- A Not assessed list covering image quality, mobile layout, page speed and how variant selection behaves
This is an illustrative scenario showing how to fill in the prompt and what to look for. It is not a recorded model output. Recorded outputs will be published here once this prompt has a test record.
Testing and editorial review
Status: Untested. Reviewed for structure and clarity by the editorial team. No recorded model test yet.
No test record has been published for this prompt yet. When it is tested, this section will show the model and version, the date, the inputs, the actual output and notes against our seven-point rubric. We do not publish numerical scores.
| Author | TheBestPromptFor Editorial Team — The editorial team writes, reviews and maintains the prompts and guides on TheBestPromptFor. Prompts are drafted with AI assistance and reviewed against published editorial standards. |
|---|---|
| Published | November 15, 2026 |
| Last reviewed | November 15, 2026 |
| Last tested | Not yet tested |
| Version history |
|
Limitations and things to check
- This is a content audit. Layout, imagery, speed and checkout behavior need a person on the live site, or separate tools.
- Severity ratings are the model's judgment, not measurements. Your analytics and customer feedback should override them.
- It cannot tell you why visitors leave, only where the page fails to answer likely questions. User testing or session recordings fill that gap.
- Recommended fixes are untested. Measure the effect of changes before rolling them out across the catalog.
Alternative prompts
E-commerce
Product Description Prompt
Writes a title, description, detail bullets and meta description from your product facts only, with a fact-check table and a list of what is missing.
- ChatGPT
- Claude
- Gemini
Web Design
UX Audit Prompt
Runs a heuristic review of one user flow: a step-by-step walkthrough, findings tied to usability heuristics with severity ratings, and fixes.
- ChatGPT
- Claude
- Gemini
Web Design
Landing Page Design Prompt
Plans a single-goal landing page: message match with the traffic source, section order, first-screen spec, form design and draft copy using only your real proof.
- ChatGPT
- Claude
- Gemini
SEO
On-Page SEO Review Prompt
Reviews one existing page against its target query and returns specific before-and-after edits to the title, headings, opening, coverage, links and images.
- ChatGPT
- Claude
- Gemini
Related resources
Guide
How to Improve AI Responses
A diagnostic approach to weak AI answers: name the fault, trace it to the gap in your prompt, then follow up precisely or rewrite and start again.
Guide
How to Write a Good AI Prompt
The five parts of a complete prompt (goal, context, task, constraints, output format), with a worked rewrite and a pre-send checklist.
Frequently asked questions
Can I paste a URL instead of the page content?
Only if the tool you use can browse the web, and even then pasted text is more reliable because you control exactly what the model reads. Pages built with heavy scripting, tabs or accordions are often retrieved incompletely, which leads to findings about content that is in fact present.
How do I include images in the audit?
Describe each one in a few words inside the pasted content, for example "photo 3: desk at standing height with a person for scale". The model can then judge whether the set of images answers shopper questions, though not their quality.
Should I audit every product page this way?
Start with pages that receive the most traffic or advertising spend. Many findings, such as unclear shipping costs or a missing returns summary, come from the page template and will apply across the catalog, so fixing them once improves every product.
Sources and references
- 10 Usability Heuristics for User Interface Design — Nielsen Norman Group
- Prompt engineering overview — Anthropic
Was this prompt useful?
No account needed. Do not include personal information.