Free law firm SEO checker · about 20 seconds · no sign-up
Can Google and an AI agent read your firm’s website?
Enter your homepage. We fetch it the way a crawler does and score what it ships: how much script a phone has to run, whether the images are modern, how well the page describes the firm to machines, and whether it offers the plain-text files AI systems look for. Then we place the result among 25 plaintiff-firm homepages we measured in October 2026, next to a site we built, scanned at the same moment.
Results
Your site
0/100
AstroPress reference build
A New York law firm’s site we built and operate, scanned just now
–/100
Each tick is one of the 25 plaintiff-firm homepages in our October 2026 scorecard. Red is you.
| Check | Your site | Reference build |
|---|
This scan reads the homepage HTML and the files it names; it does not run Lighthouse or read real-user Core Web Vitals, and a tag manager can load more tags than the page names. Script sizes are uncompressed. Scores use our rubric below, not an industry standard.
The scan sees the page. The Architecture Review covers the rest: who owns it, where the leads go, and what happens between the first search and a signed case.
What this measures
An agent does not see your website. It reads it.
An AI agent is software acting for a person: an assistant asked to find a lawyer, compare two firms, or check whether a firm handles a kind of case. It gets the page as data. It reads the HTML, the accessibility tree a browser builds for screen readers, any structured data the page declares about itself, and, if the site offers them, plain-text copies it can read without parsing the layout at all.
Google added a category for that reader to Lighthouse in May 2026. It is called Agentic Browsing and it is still marked as under development. This score covers the part Lighthouse does not: what the site ships before any rendering happens, and whether it describes the firm to a machine at all.
People sell this work under several names, GEO and AEO among them. Nobody can tell you what share of your prospects ask an assistant first, and we will not pretend to. What can be measured is whether the site is readable when one does. That is the floor, and it is the part a firm can fix on its own.
a person sees
A photograph, a headline, a red phone number, a reviews carousel, and a form that emails an inbox. Twelve scripts and three tags behind it.
an agent reads
- index.html script weight, blocking scripts, image formats
- accessibility tree named buttons, labelled fields, a structure to follow
- JSON-LD · LegalService what the firm is, where it practises, what it handles
- llms.txt · index.md a plain-text index and a Markdown copy, read without parsing the layout
Where your site lands
Twenty-five plaintiff firms, one strip.
On October 16, 2026 we scored the homepages of 25 plaintiff-side firms across seven markets with this exact scan. The median was 42. The median page weighed 2.2 MB on a phone. Twelve named more than 400 KB of script. Three sent a Content Security Policy and none of the three restricted scripts.
The reference build, a New York firm's site we built and operate, scored 84 the same day. When you run the scan, your site appears on this strip in red.
Read the full scorecardWhy it matters
Two firms, same fundamentals. One difference.
Same practice, same market, same years at the bar, the same pages saying the same things. Google has said what happens when the content is a draw: “in cases where there are multiple pages that have similar content, page experience becomes much more important for visibility in Search.” Its current documentation is shorter: “Core Web Vitals are used by our ranking systems.”
The median homepage in the field takes 5.9 seconds to paint its largest element on a phone, against Google's threshold of 2.5. The firm with the lighter site is not the better firm. On Google's own account, it is the one of the two that passed the test.
Google Search Central, May 2020, and its current page experience documentation. We did not measure rankings for any site.
Firm Athe median homepage in the field
vsFirm Bthe reference build, same day
held equalpractice · market · years at the bar · pages · reviews
Readiness scorehigher is better
Homepage weight on a phonelower is better
Script named in the pagelower is better
Lighthouse mobile performancehigher is better
“…in cases where there are multiple pages that have similar content, page experience becomes much more important for visibility in Search.”
Google Search Central, Evaluating page experience for a better web, May 2020
The rubric
How the score is built
Ten checks, 100 points, every threshold published. The weights reflect what we see costing firms most on phones: script weight and thin structured data. Disagree with a weight and the measurement underneath it is still yours to use.
The scan does not run Lighthouse or read real-user Core Web Vitals, and a tag manager can load more tags than the page names. Script sizes are uncompressed. What agent-ready means in practice is in the briefing.
- 20JavaScript weightUnder 100 KB of script passes; over 400 KB fails.
- 15Structured data depthJSON-LD that names a legal-service type and describes it in depth.
- 10Modern image formatsAt least 80% of raster images served as AVIF or WebP.
- 10llms.txtA plain-text index of the site written for AI systems.
- 10Markdown copy of the pageA clean text version an agent can read without parsing layout.
- 10Content Security PolicyA policy that actually restricts which scripts may run.
- 10Static asset cachingScripts and styles cacheable for a month or more.
- 7Third-party trackingOne known tag or fewer named in the page.
- 5Render-blocking scriptsNo synchronous scripts in the head.
- 3AI crawler accessThe major AI crawlers are not blocked in robots.txt.
Questions
What firms ask before they run it
Is this the same as Lighthouse or PageSpeed Insights?
No. Lighthouse loads the page in a simulated browser and scores performance, accessibility, best practices and SEO, plus its newer Agentic Browsing checks. This scan reads the homepage HTML and the files it names, before any rendering, and scores ten things an AI agent or crawler meets first: script weight, blocking scripts, image formats, structured data, llms.txt, a Markdown copy, the security policy, caching, tracking and crawler access. Run both. They answer different questions.
What do people mean by GEO or AEO, and does this score measure it?
Generative engine optimization and answer engine optimization are names for the work of getting a firm cited when an AI assistant answers a question. Nobody can tell you what share of your prospects ask an assistant first, and this score does not try. It measures something narrower and checkable: whether the site is readable when an assistant or agent arrives. That is the floor of any GEO effort, and the part a firm can fix on its own.
Why does JavaScript weight count for 20 of the 100 points?
Because it is the cost we see most often. In October 2026 the median plaintiff-firm homepage named 333 KB of script, and twelve of the 25 we measured named more than 400 KB (Wolfe Services website scorecard, 2026). Script is work a phone has to do before the text settles, and work an agent has to wait through. Under 100 KB passes; over 400 KB fails.
What is llms.txt?
A plain-text file at the root of a site that tells AI systems what the site is and where its important pages are, as proposed at llmstxt.org. Eleven of the 25 homepages we measured had one, and two of those failed the basic checks Lighthouse applies to it, so the file alone is not the point. It has to describe the firm accurately and point at pages that are worth reading.
What is a Markdown copy of a page?
A clean text version of the page that an agent can fetch instead of the HTML: the headings, the paragraphs and the links, with none of the layout. None of the 25 homepages we measured offered one. This site publishes one beside every page; add .md to almost any address here to see it.
Does a good score mean more cases?
No. Nothing here measures rankings, calls or signed cases, and we did not try to connect them. A readable site is a precondition, not a result. The Architecture Review is where the path from the first search to the signed case gets examined: who owns the site, where the leads go, and what happens in the first hour after one arrives.
After the score
The scan sees the page. The review covers the rest.
A readable site is a precondition, not a result. Nothing on this page measures rankings, calls or signed matters, and we do not pretend it does. The next step depends on what the result tells you.
Request an Architecture Review- 01
Read the check-by-check result
Every row shows the number we measured, the grade and the threshold it was graded against, next to the same check on the reference build.
- 02
Hand it to whoever runs the site
The fixes that move the score most are script weight, structured data and a Content Security Policy, in that order. None of them needs a redesign.
- 03
If the problem is bigger than the page
The Architecture Review covers what the scan cannot see: who owns the site, where the leads go, and what happens between the first search and a signed case.