A useful review
has clear limits.

Agree the scope first

Each review covers up to three agreed public pages at desktop and mobile sizes. We identify the main visitor task before testing. Password-protected areas, purchases, submissions to your customers, and account changes require separate permission and are not included by default.

Show the work

A delivered finding should record the page, browser and viewport, steps to reproduce, expected and observed behavior, and a supporting screenshot where useful. Recommendations should include a way to check the fix. Unverified observations are labeled as such rather than presented as facts.

Prioritize by visitor impact

• High: an issue that blocks or seriously disrupts a key visitor task.
• Medium: friction that makes a task harder or removes useful context.
• Low: a refinement with less immediate impact on the agreed journey.

Priority is a reasoned judgment, not a revenue prediction. The report highlights three recommended starting points.

Use AI carefully

The planned workflow may use Claude to help organize observations and draft report wording. A person verifies findings and recommendations. There is no live Claude integration or automated scanning on this website. Any sharing of private material with an AI service would require an agreed process first.

Know what this is not

A focused review is not a penetration test, legal review, comprehensive accessibility audit, SEO campaign, or certification of compliance. Performance observations depend on the test conditions and are not a guarantee. Search positions, conversion gains and complete defect detection cannot be promised.

Fixes are a separate decision

No changes are made to your website as part of the review. Any implementation work requires a separate scope, price and approval. Pilot pricing, availability and turnaround will be confirmed before accepting work.