The formula
Ten areas. Each scored 0–100 on its own merits, then multiplied by a fixed weight and added together. The weights reflect how much each area affects whether a stranger becomes a paying customer, not how interesting the area is to talk about.
Overall = Σ (area score × area weight) ÷ 100
| Area | Weight | |
|---|---|---|
| Homepage conversion | 12% | |
| Quote path & lead capture | 12% | |
| Brand clarity | 11% | |
| Trust & review leverage | 11% | |
| Local discovery & GBP | 11% | |
| Service packaging | 10% | |
| AI Readiness & GEO | 10% | |
| Mobile & contact path | 9% | |
| Social proof & engagement | 7% | |
| Measurement & follow-up | 7% | |
| Total | 100% |
Score bands
The same five bands apply to each area score and to the overall number.
Working as intended. Polish only.
Solid foundation, real gaps remain.
Half-built. Costing you now.
Actively leaking leads.
Fix before spending anything on traffic.
What each area covers
Findings vary by business, so this is not an exhaustive list. It is the backbone of every audit.
Homepage conversion · 12%
- First screen names what you sell, who it is for, and where you work
- A primary action visible without scrolling, on desktop and on a phone
- Proof placed anywhere near that primary action
- Number of competing calls-to-action fighting each other above the fold
- Page weight and load time. A slow first paint is a silent conversion tax
Quote path & lead capture · 12%
- Clicks from homepage to a submitted enquiry
- Form field count, and how many are truly required
- Does the form stay on your domain, or hand the visitor to a third party
- Any indication of what happens after they hit submit
- Confirmation emails. We submit the form ourselves and wait for one
Brand clarity · 11%
- The five-second test: what does a stranger think you do?
- Page title and meta description matching the real business
- Service area stated in words, not implied by a map
- Copy that describes the customer's problem, or only lists your services
- Consistent naming across your site, your listings and your social profiles
Trust & review leverage · 11%
- Review count, average rating, and the date of the most recent one
- Replies to reviews, especially the bad ones
- Proof on the site itself, and how close it sits to the ask
- Real photos of real work, or stock imagery
- Named people, licences, guarantees and warranties
Local discovery & Google Business Profile · 11%
- A profile that is claimed, verified and complete
- Category accuracy, primary and secondary
- Photo count and recency, services listed with descriptions, and recent posts
- Name, address and phone matching across the site and the major directories
- Your ranking for your main service plus your city
Service packaging · 10%
- A real page per service, or everything buried in one paragraph
- Scope and boundaries stated plainly: what is and is not included
- Any pricing signal at all, even a starting range
- Enough detail for a buyer to pick the right service themselves
AI Readiness & GEO · 10%
This is the newest area and the one most businesses have never had checked. Much of it is directly testable:
- Crawler access test. We request your site as ClaudeBot, GPTBot, PerplexityBot, OAI-SearchBot, Google-Extended, and Googlebot, and record the HTTP status each one gets back. A 403 means that assistant cannot read you at all.
- robots.txt and llms.txt. What your files really permit, including managed rules your host may have added without telling you.
- Structured data. Does LocalBusiness or Organization schema exist, and are the facts inside it correct and complete?
- Retrieval test. We ask several AI assistants who you are and what you do, and record whether the answer is accurate. Wrong hours or a dead phone number in an AI answer is worse than not appearing at all.
- Invisible content. Hours, phone numbers, prices, or service lists that exist only inside images, where no crawler can read them.
- Answerability. Do your pages answer the questions buyers really ask, in language an assistant can quote?
Mobile & contact path · 9%
- A phone number that is a tap-to-call link, not plain text
- That number reachable without scrolling on a phone
- Tap target sizes, text legibility and horizontal scroll problems
- Mobile load time on a realistic connection
Social proof & engagement · 7%
- Which platforms have a real presence versus an abandoned one
- Last post date, and how consistent the posting is
- Followers who engage, or an account that is quietly dormant
- Profiles that link back to the site and agree with it
Where a platform blocks automated checks behind a login, we say the check was blocked rather than guessing. An unverifiable profile is reported as unverified, not as absent.
Measurement & follow-up · 7%
- Analytics and conversion tracking installed at all
- Form submissions and calls recorded as events
- An automatic response going out when a lead arrives
- A response-time expectation stated publicly
- Could you answer "where did last month's leads come from?"
Evidence standards, and the limits of this
Every score ships with its evidence
No score appears without something behind it: a screenshot, an HTTP status code, a field count, a quoted line of your own copy. If we claim your contact form has too much friction, the report shows the form.
You are welcome to disagree. Several findings in past audits changed after the owner explained context we could not see from outside.
What we cannot see
This audit works from public surfaces plus whatever you tell us. Without account access we cannot see your ad spend, your analytics history, your CRM, or your close rate.
Where something is unverifiable, the report says so plainly. It does not get scored as a failure just because a login blocked the check.
What we deliberately do not do
- ✕Predict a revenue number. Anyone who tells you an audit will produce "an extra $40,000" is guessing at your business.
- ✕Score an area we could not check, in order to fill out the page.
- ✕Adjust weights per client so the final number looks more flattering.
- ✕Recommend a service we sell as the fix for a problem that has a free fix.
See the rubric applied to a real business
We ran this exact methodology on our own site and published the result unedited. We scored 61 out of 100, including a 34 on social proof and a 61 on AI Readiness, which is not a comfortable number for a company that sells AI Readiness.