AI Website Audit Prompts: Seven Checks Before a Redesign

Website owners and conversion teams can run AI website audit prompts before a redesign to find the most costly clarity and decision problems in the page they already have. This seven-part pack moves from a visitor’s first impression through copy, proof, calls to action, mobile friction, search structure, and pricing—then asks the team to fix one evidenced issue at a time.

Pack Overview

Use the prompts in order when the immediate goal is a clearer existing website, not a wholesale rewrite. Provide the relevant page URL, copied text, screenshots, or verified page data as the input; retain a human owner for brand claims, accessibility, analytics, and release decisions.

Strategic Deployment Guide

Model fit: Claude is a practical fit for these text-led review prompts because the source pack is written for Claude. Start by pasting the specific page material named in each step; do not ask any model to invent analytics, user behavior, legal claims, or technical crawl evidence it cannot see.

Prompt 1: First Impression Audit

• Target: test what a new visitor understands before scrolling.
• Input: homepage URL, screenshot, or complete homepage text.
• Model fit: use Claude with the actual page material.
• Expected output: a concise statement of perceived offer, audience, and confusion points.
• Quality check: confirm the flagged issue is visible in the supplied material before changing copy.

Act as a conversion rate expert. Based on this homepage, tell me exactly what a first-time visitor understands in the first 5 seconds and what confuses them. Be blunt.

Prompt 2: Copy & Messaging Clarity Audit

• Target: identify business-centred copy that leaves the reader’s outcome unclear.
• Input: homepage or landing-page copy.
• Model fit: use Claude for a text-level critique, then have an editor check the rewrites against approved claims.
• Expected output: the weakest reader-misaligned lines and three rewrites.
• Quality check: preserve factual claims, product terminology, and legal wording where required.

Review this copy like a direct-response copywriter. Point out every sentence that talks about the business instead of the reader, and rewrite the 3 weakest lines.

Prompt 3: Trust & Credibility Signals Audit

• Target: surface missing proof a skeptical buyer would reasonably need.
• Input: About-page or homepage copy, plus only verified proof assets.
• Model fit: use Claude to organize evidence gaps; do not fabricate reviews, credentials, guarantees, or results.
• Expected output: a prioritized list of missing trust signals.
• Quality check: every added signal must be verifiable and approved by the business owner.

Audit this page for trust signals social proof, credentials, guarantees, specificity. Tell me what's missing that a skeptical visitor would want to see before buying.

Prompt 4: Call-to-Action Audit

• Target: reduce vague or competing calls to action.
• Input: every CTA label and its nearby headline or context.
• Model fit: use Claude for language alternatives, then validate with actual conversion evidence.
• Expected output: three revised CTAs and an explanation of which ones lack clarity or appropriate urgency.
• Quality check: check button meaning, destination, and accessibility label after publishing.

Review these CTAs. Tell me which ones are vague, which create urgency without being pushy, and rewrite the weakest 3.

Prompt 5: Mobile & UX Friction Audit

• Target: find the most likely mobile drop-off point before making many changes.
• Input: page structure, menu, and checkout or booking flow, ideally with mobile screenshots or test observations.
• Model fit: use Claude to prioritize supplied flow evidence; a browser test is still required for actual mobile behavior.
• Expected output: a ranked set of likely friction points and the simplest fix for each.
• Quality check: test the first fix on a real mobile viewport before addressing lower-priority suggestions.

Based on this structure, act as a UX auditor. Identify where a mobile visitor is most likely to get confused or drop off, and suggest the simplest fix for each.

Prompt 6: SEO & Structure Audit

• Target: test whether the title and heading hierarchy match the intended search task.
• Input: homepage title, meta description, H1, and H2s.
• Model fit: use Claude for a semantic clarity review; verify demand and indexing separately with search data.
• Expected output: a diagnosis plus three improved title-and-heading directions.
• Quality check: do not replace live SEO fields until the target query, evidence, and page purpose agree.

Review this page structure for SEO clarity. Tell me if the title and headings actually match search intent, and suggest 3 improved versions.

Prompt 7: Pricing Page Psychology Audit

• Target: check whether plan presentation makes a confident choice easier.
• Input: pricing-page copy and plan names, with verified prices and terms.
• Model fit: use Claude to name decision-friction patterns; pricing, consumer-protection, and experiment decisions require human approval.
• Expected output: one proposed change that makes the default choice clearer.
• Quality check: validate any change against real pricing policy and test it before adding plans or discounts.

Act as a pricing psychology expert. Tell me if this pricing page reduces decision fatigue or increases it, and suggest one change that would make the default choice obvious.

Selection Logic

• Start with Prompt 1 when visitors may not understand the offer or audience.
• Use Prompts 2–4 when the offer is understood but copy, proof, or the next action feels weak.
• Use Prompt 5 when a real mobile flow or session evidence points to friction.
• Use Prompt 6 only with the actual title and headings, then verify any search-facing change separately.
• Use Prompt 7 when buyers hesitate between plans or fail to see a sensible default.

Implementation Steps

• Paste one page’s exact copy, visible CTA labels, headings, plan names, and approved screenshots or behavior evidence into the matching audit prompt.
• Run Prompt 1 first, then paste the discovered audit issue into only the next prompt that addresses it.
• Record the audit observation, the supplied evidence, the proposed change, and a human owner.
• Apply one small, reversible CTA, mobile, or SEO change and test it in a real browser, analytics tool, or reviewer workflow rather than treating model output as proof.
• Repeat the relevant audit prompt only after documenting whether the prior change clarified the page.

Use Cases

• SaaS founders reviewing a homepage: clarify the value proposition and the primary trial or demo action before a redesign brief.
• Service-business marketers reviewing a booking page: check whether claims, proof, and booking prompts answer a cautious buyer’s questions.
• Ecommerce managers reviewing a category or pricing page: reduce choice friction while keeping product and pricing facts approved.
• Agency strategists running a discovery audit: create a structured discussion of page evidence and priorities before proposing a scope of work.

Why These Prompts Work

The pack separates seven common website decisions instead of asking for an unbounded “full audit.” Each prompt narrows the input, asks for a specific diagnostic outcome, and ends with a small next action. That makes it easier to reject unsupported suggestions, preserve facts, and prioritize a reversible fix before expanding the scope.

Troubleshooting & Optimization

• The feedback is generic: supply the exact page text, CTA context, or pricing details rather than a broad site description.
• The model invents proof or metrics: remove the invented claim and provide verified evidence only; lack of proof is itself an audit finding.
• Too many changes are suggested: use the selection logic to choose one observed failure and test the smallest change first.
• SEO suggestions conflict with intent: preserve the documented user task and confirm demand before changing live titles or headings.

FAQ

• Q: What are AI website audit prompts?
  A: They are targeted instructions that ask an AI model to assess supplied page evidence for a specific clarity, trust, conversion, UX, search, or pricing question.
• Q: Can these prompts replace analytics or usability testing?
  A: No. They help structure a review of the information you provide. Validate behavior, conversion, accessibility, and search performance with the appropriate real evidence.
• Q: Which prompt should I run first?
  A: Start with the first-impression audit. If the offer is unclear, later copy, CTA, or pricing changes will be harder to judge.
• Q: Should I apply every recommendation at once?
  A: No. Make one evidence-backed, reversible change, record the outcome, and then move to the next question.

Use this prompt pack when an existing site needs a focused diagnostic before a redesign. Explore Prompt Engineering Guides and Marketing & Growth for adjacent systems.

Follow @bigprompt for more reusable AI workflows and prompt systems.

Related Big Prompt Hub pages:

AI Web Design Skills That Improve Homepage Flow and CTA Structure

Architect a Strategic Website Plan Prompt for Business Growth

SEO Copywriting Prompt: Business Website Content Generation

Figma to React Agent Workflow for Production Components

Agent Skills Evaluation Workflow: Evidence Before Installation

Big Prompt Hub Review

Use this pack as a pre-redesign decision aid, not an automatic website grader. Start with supplied evidence, keep the scope to one page and one observed problem, then let an accountable owner verify claims, behavior, accessibility, search effects, and pricing before shipping a change.