AI Disclosure
Audit&Fix uses artificial intelligence in two places: as part of our website audit process, and in the chat widget on this site. This page explains exactly how we use it, what it can and cannot do, and how we keep a human in the loop.
What AI Does in Our Audit Process
When you submit your website for an audit, our system takes screenshots of your key pages and feeds them, along with your page copy and structure, into an AI analysis engine. The AI examines layout, calls to action, trust signals, mobile experience, page speed indicators, and dozens of other conversion factors. It produces a scored breakdown of findings and draft recommendations.
Think of the AI as a very fast, very thorough first analyst. It catches things a human eye might skim past, and it applies a consistent scoring framework across every audit. But it does not have the final say. A member of our team reviews every report before it reaches you, adjusting scores, adding context, and removing anything that does not apply to your specific business.
The Chat Widget (Marcus)
The chat widget in the bottom-right corner of this site — labelled "Marcus / Virtual Assistant" — is an automated AI assistant, not a human. It's powered by ContactReplyAI, our own automated reply product, which we're using on this site to dogfood our own service.
What Marcus can do: answer questions about our pricing, how the audit works, what you get, turnaround times, and point you to the right page on the site.
What Marcus will not do: make binding commitments, quote custom pricing outside published tiers, give regulated professional advice (legal, medical, financial), or pretend to be human. If you sincerely ask "is this AI?" Marcus will confirm.
If you'd rather talk to a person, use our contact form — handled by a real person, not by AI. Contact-form submissions go directly to the team inbox.
What AI Does Not Do
The AI does not make decisions on its own about your audit. It does not approve, reject, or send any report without human review first.
It does not have access to your personal information beyond the website URL you submitted, the publicly visible content on that site, and (for chat) the messages you type into the widget. It does not read your emails, access your analytics, or connect to any of your third-party accounts.
The AI does not retain memory of your audit between sessions. Each analysis is independent. Chat transcripts are retained for service-quality review but are not used to train the AI models we rely on. Your audit data is not used to train the AI either.
Which AI Systems We Use
We use Claude, built by Anthropic, accessed through their API. Anthropic is a safety-focused AI company based in San Francisco. Under Anthropic's commercial API terms, data sent through the API is not used to train their models. You can read Anthropic's usage policy and privacy documentation at anthropic.com for full details.
We do not use AI from multiple competing providers. Claude is the only model involved in generating audit content.
Human Oversight
Every audit report is reviewed by a team member before delivery. The AI produces a draft with scores and recommendations. A human then reads that draft in full, checks the scores against what is actually on your site, adjusts findings where the AI has missed context, and removes any recommendation that does not make sense for your industry or business model.
AI scores are a starting point, not a final verdict. If the AI flags something as a problem but our reviewer disagrees after looking at your site, the finding gets amended or removed. The report you receive reflects human judgement informed by AI analysis, not the other way around.
Accuracy and Limitations
AI is good at pattern recognition and applying known best practices consistently. It is less good at understanding the specific context of your business, your customers, or your industry norms.
For example, the AI might flag a missing phone number as a trust issue on a service page, but your business might operate entirely online with no phone support by design. Our human review catches most of these, but not every edge case.
Recommendations in our reports are based on general conversion rate optimisation best practices. They are not guarantees of specific results. Implementing a recommendation may improve your conversion rate, but outcomes depend on your traffic, your offer, your market, and many factors outside the scope of a website audit.
If something in your report feels off or does not apply to your situation, we want to hear about it.
Your Rights
You can request a human re-review of any finding in your report at no extra charge. If you believe the AI got something wrong, or if a recommendation does not account for something specific to your business, use our contact form with the finding in question and we will take another look.
You can also ask us to explain how any particular score or recommendation was generated. We are happy to walk you through the process.
For questions specifically about our use of AI, use our contact form with the subject line AI Disclosure Query and we will respond within two business days.
Changes to This Policy
We may update this page as our tools and processes evolve. When we make meaningful changes to how AI is used in our audit process, we will update this page and note the date above. We will not reduce the level of human oversight described here without updating this disclosure first.