This is the editable version: a real SEO audit as a .docx you can open, rewrite and send under your own name. It opens in Google Docs as well as Word, and every section is plain styled text rather than a locked layout, so replacing the content does not break the document.
This is the file to take if you are writing an audit for a client this week and want a structure that has already survived client questions.
Download the Word report
The full audit of pearllemon.com — 72/100 across six pillars, 14 findings. No email, no signup; the file is built in your browser when you click.
Same report, other formats
What to replace and what to keep
Three things in this document are worth keeping when you swap in your own client's data, because they are what stop an audit reading like a sales pitch.
Keep the evidence line under each finding. "Slow site" is an opinion; "33 render-blocking scripts, 45 external stylesheets" is a fact and also a scope.
Keep the passed checks. An audit listing only failures reads as a pitch, and the client will read it that way.
Keep the weighted score arithmetic. It converts a defensive conversation about the number into a scoping conversation about a pillar.
Rewriting it for a client in about an hour
The order that works: run the audit on the client's domain first so you have the real numbers in front of you, then work down this document replacing findings in place. Delete findings that do not apply rather than rewriting them into something vague — a short honest audit is worth more than a padded one.
Last, delete our footer. The document has no other branding in it.
Why Word rather than a design file
Audits written in design tools look better and get sent later. A .docx can be edited by whoever is free, tracked-changed by a second pair of eyes, and exported to PDF at the end, which is the only format the client actually receives.
If you want the finished-looking version instead, the PDF of this same report is one click away, and the Excel workbook holds the findings as a filterable table.
If you would rather not rewrite it each time
Rewriting a template per prospect is fine for one or two. Past that, running the audit and exporting a branded document directly is faster, and the findings are the client's rather than ours with the names changed.
Score breakdown
pearllemon.com scored 72/100 when we audited it on . The overall figure is a weighted average of six pillars, so you can always reconstruct it by hand. The benchmark column is the average across 126 sites we have audited.
3 critical issues, 7 warnings and 4 passed checks. Each one records what was measured, not just a verdict — that evidence line is what makes an audit defensible when a client pushes back.
Critical (3)
No agent interfaces published
AI & E-E-A-T readiness
What we measured:
Agent-interface score 5/100. No /.well-known/agents.json, no WebMCP script tags, no AGENTS.md.
Recommendation:
Publish agents.json declaring your key user flows, add AGENTS.md, and expose WebMCP tools so AI agents can act without scraping the DOM.
Authorship and dates are invisible to machines
AI & E-E-A-T readiness
What we measured:
Authoritativeness score 10/100. No schema.author, no machine-readable <time datetime=…>, 0 authoritative outbound citations.
Recommendation:
Add a visible author byline wired to schema.author, surface a publish/updated date in <time>, and cite two to three .gov/.edu/.org sources.
Security header coverage is incomplete
Security
What we measured:
Security pillar scored 48/100 against the header and transport checks.
Recommendation:
Add Content-Security-Policy, Strict-Transport-Security, X-Content-Type-Options, Referrer-Policy and Permissions-Policy at the edge.
Warning (7)
33 render-blocking scripts and 45 external stylesheets
Performance
What we measured:
111 external resources in total; HTML document weighs 688 KB before assets.
Recommendation:
Defer non-critical JavaScript, combine and inline critical CSS, and drop unused plugin stylesheets. This is the single biggest lever on LCP here.
No images are lazy-loaded
Performance
What we measured:
0% of 58 images use loading="lazy"; only 41% use a modern format.
Recommendation:
Add loading="lazy" below the fold and serve WebP/AVIF for the remaining 59% of images.
16 form inputs have no associated label
Accessibility
What we measured:
16 unlabelled inputs, 14 links with no discernible text, no <main> landmark.
Recommendation:
Associate every input with a <label>, give icon-only links accessible names, and wrap the primary content in <main>.
Only one question-style H2
Content
What we measured:
AI citation score 55/100; one question-format heading across 14 H2s. Answer-first opening paragraph was detected.
Recommendation:
Convert two to three H2s into the questions people actually search ("What is…", "How do I…") — this is what answer engines quote.
Primary call to action sits far below the fold
Content
What we measured:
First CTA appears 7,268 px down the page. Conversion-readiness grade 88/100 otherwise.
Recommendation:
Move a primary CTA into the first viewport.
Inconsistent NAP formatting
Technical SEO
What we measured:
Two different phone-number formats detected across the page and schema.
Recommendation:
Normalise every phone number to a single E.164 format in both copy and LocalBusiness schema.
No web app manifest
Technical SEO
What we measured:
PWA score 3. Favicon and apple-touch-icon present; manifest and theme-color absent.
Recommendation:
Add a manifest.json with icons and a theme colour.
Passed (4)
Crawlability is clean
Technical SEO
What we measured:
robots.txt present, XML sitemap available, canonical declared, all 18 priority AI bots allowed.
Recommendation:
No action needed. Re-check after any hosting or CDN change.
Structured data is well covered
Technical SEO
What we measured:
Structured-grounding score 90/100: Organization, LocalBusiness, Article-class and FAQPage schema all present.
Recommendation:
No action needed. Keep FAQPage in sync with the visible copy.
Trust and social proof are strong
Content
What we measured:
Testimonials, case studies, client logos, trust badges and a Trustpilot profile all detected.
Recommendation:
No action needed.
Image alt coverage is near complete
Accessibility
What we measured:
57 of 58 images carry alt text; a skip link and <nav> landmark are present.
Recommendation:
Fix the single remaining image and this check is clean.
Raw measurements
The underlying numbers the scores were derived from. A report that shows only scores is impossible to check; a report that shows measurements can be re-run by anyone.
Measurements recorded during the audit of pearllemon.com
Metric
Value
Words of body copy
2,780
Heading structure
1 × H1, 14 × H2, 13 × H3
Internal links
134
External links
15
Images
58 (1 missing alt text)
Modern image formats
41% of images
Lazy-loaded images
0% of images
Render-blocking scripts
33
External CSS files
45
Total external resources
111
HTML document size
688 KB
Readability (Flesch)
52 — grade 9.9, "difficult"
AI crawlers allowed
18 of 18 priority bots
How 72/100 compares
Every completed audit in the LemAudit database that carries a full pillar breakdown. Sites were submitted by agencies and prospects, so the sample skews towards small and mid-sized business websites rather than enterprise.
The average site in that sample scores 65/100. Two patterns hold almost universally: accessibility and technical fundamentals are in decent shape, while content depth and machine-readable authorship are not. Content is the weakest pillar overall at 31/100.
Most common worst-issue across 126 audited sites
Most common worst issue
Sites affected
No agent interfaces detected
104 of 126
Very thin content
17 of 126
No WebMCP tools exposed
12 of 126
Duplicate <title> tags
5 of 126
Missing H1 tag
2 of 126
robots.txt blocks all crawlers
2 of 126
Frequently asked questions
Do I need Microsoft Word to open it?
No. The .docx opens in Google Docs, LibreOffice and Pages. It uses plain paragraph and heading styles, so nothing shifts between applications.
Is this a blank template or a filled-in report?
Filled in, with a real audit. A blank template does not tell you what belongs in each section, which is the part most people are actually looking for. Delete the data and you have the blank version.
Can I remove your branding?
Yes. There is one footer line and no other mark in the document. Delete it and the report is yours.
Is there a Markdown version for AI tools?
Yes — the same report exports as Markdown, which is the format to use if you are feeding it into a language model or a documentation site.
How long does it take to adapt for a client?
About an hour if you already have the audit data for their domain. Most of that time goes on deleting findings that do not apply rather than writing new ones.
