
Last updated September 2026
Yes, AI can build a law firm website, and the design will probably look fine.
What decides whether it protects your firm is the hour after the tool finishes, when someone has to audit accessibility, check it against your bar’s advertising rules and write Spanish your clients can use, and all three jobs land on the attorney, not the tool.
You should know where I stand before reading further.
I build websites for immigration firms and I build them with AI assistance, including the service page that sells them and the six concept builds shown on it, so I have no reason to argue against the tool.
Every number below links to its source, and several of them cut against my own sales pitch, which is the best reason I can offer you to trust the rest.
What AI website builders do well in 2026
This part is real, and pretending otherwise would waste your time.
Current builders produce responsive layouts on the same frameworks agencies use, publish to a live URL with SSL in minutes, and write service copy that reads fine on a first pass.
Here is a number I did not expect to find.
In the 2026 WebAIM Million report, which tested the home pages of the top one million sites in February 2026, Squarespace pages averaged 33.0 accessibility errors and Wix averaged 33.3, both roughly 41% below the overall average, while WordPress averaged 52.8.
The easy builders beat the platform most agencies use, mine included, so if anyone tells you simple tools produce worse websites, that claim does not hold up against the largest dataset available.
The weak point sits somewhere else, and the rest of this piece is about where.
Your Next Client Is Searching for You Right Now
Immigration clients are online every day, looking for a law firm they can trust. I help immigration attorneys show up in those searches, build credibility before the first call, and turn website visitors into booked consultations. See how I do it and what it costs for your firm.
The web got measurably less accessible in 2026
That same report found that 95.9% of home pages had detected WCAG 2 failures, up from 94.8% in 2025, which reverses six straight years of small improvement.
The average page carried 56.1 detectable errors, a 10.1% increase over the year before.
Six failure types account for 96% of everything detected, and they have been the same six for seven years running:
| Failure type | Share of home pages |
| Low contrast text | 83.9% |
| Missing alternative text on images | 53.1% |
| Missing form input labels | 51% |
| Empty links | 46.3% |
| Empty buttons | 30.6% |
| Missing document language | 13.5% |
One third of all form inputs, 33.1%, were not properly labeled, and on a law firm site the form is your intake, so that particular failure costs you consultations directly.
WebAIM offers its own explanation for the reversal, and it belongs in this article.
In the report’s conclusion, they point to broader shifts in web development, naming heavier reliance on third-party frameworks and libraries alongside automated and AI-assisted coding practices, and they use the term vibe coding directly.
Complexity backs that up, since the average home page carried 1,437 elements in February 2026, a 22.5% jump in a single year.
Generated code tends to come out heavier than hand-built code, and heavier pages fail accessibility checks more often, which is why somebody has to test the output before it goes live.
How much litigation risk is attached, by state
Most marketing pages stop at the national number and let you assume the worst.
The state-level data tells a different story, and if you practice in California it points almost the opposite way.
Seyfarth Shaw counted 3,117 federal website accessibility lawsuits in 2025, 665 more than in 2024 and a 27% increase, which made up 36% of the 8,667 federal ADA Title III filings that year.
Here is where those website cases were filed:
| State | Federal website accessibility lawsuits, 2025 |
| New York | 1,021 |
| Florida | 961 |
| Illinois | 585 |
| Minnesota | 162 |
| Pennsylvania | 137 |
| Missouri | 86 |
| Massachusetts | 57 |
| New Jersey | 22 |
| California | 4 |
California had four of these cases in the entire year.
Seyfarth’s explanation is that California’s federal and state appellate courts have concluded online-only businesses are not covered by the ADA, which makes viable filings against them very difficult in that state, although they note that demand letters and state court cases continue, particularly against businesses with a physical location, and their federal count does not include those.
So if you practice in California, the federal litigation argument for accessibility is weak, and you should discount anyone who leads with it, including me.
If you practice in New York, Florida or Illinois, the filing counts run more than a hundred times higher, and the question deserves a conversation with your own counsel rather than with a web designer.
I am not a lawyer and none of this is legal advice, only a count of filings published by a firm that tracks them, with their methodology available at the link above.
The better reason to care
Take the lawsuits out of the picture and the argument gets simpler and harder to dismiss.
A prospective client reading your site at eleven at night, on a phone, worried about a case, is the person an unlabeled form fails first, and someone using a screen reader cannot submit an intake form whose fields carry no labels at all.
You lose that consultation quietly, with no complaint and no letter, since the person simply leaves and calls the next firm on the list.
I made the same point about what happens in the first seconds on an immigration firm’s website, and it applies here with more force.
Your bar rules did not change because a tool wrote the page
The American Bar Association issued Formal Opinion 512 on generative AI tools on July 29, 2024, its first formal opinion on the subject.
It addresses legal practice in general rather than marketing, and it is not binding, but the principle carries over, since using AI moves none of your existing duties and the supervising lawyer stays responsible for what goes out under the firm’s name.
Advertising exposure runs through Rule 7.1, which prohibits false or misleading communications about a lawyer’s services.
Generated copy tends to drift toward reassuring, outcome-flavored sentences, since that kind of writing is what the training data rewards, and those are exactly the sentences Rule 7.1 was written for.
States are moving on this too.
California’s State Bar put proposed amendments to the Rules of Professional Conduct related to artificial intelligence out for public comment with a deadline of May 4, 2026, and if those land in rule text instead of guidance, the obligations stop being advisory.
The archiving requirement almost nobody mentions
This is the detail I did not know before writing this piece, and the one worth forwarding to a partner.
Under New York Rule 7.1(k), every advertisement must be pre-approved by the lawyer or firm, with a copy retained at least three years after initial dissemination, and computer-accessed communications retained at least one year.
The rule goes further for websites.
The contents of any covered website must be preserved at initial publication, at any major redesign, at any meaningful and extensive content change, and in no event less often than once every 90 days, which means a copy every quarter even when nothing on the site has changed.
A generated site with nobody assigned to it produces no archive at all.
If a question surfaces eighteen months from now about a sentence that sat on your home page last spring, the issue is whether you can produce that page as it existed, and a firm advertising across state lines has to meet the strictest rule that applies to it rather than assume a single baseline.
The Spanish pages are where this bites hardest
Most AI builders will generate and translate across languages on request, which works well enough for a restaurant menu.
For an immigration practice, the Spanish page is where a frightened person decides whether you understand their situation, so the stakes are of a different order.
The WebAIM data has something specific to say here. Spanish-language home pages averaged 64.3 detected errors, 14.7% above the overall average, while English-language pages averaged 46.0, which is 18% below it.
The Spanish web is measurably worse built than the English web.
For your firm, that means the version of your site used by the clients who need the most help is statistically the one most likely to fail them, and it is also the version you are least able to proofread if you do not read Spanish yourself.
Machine translation adds a second problem.
It renders case types literally, so pages describe matters in phrasing no client would search for or recognize, and auto-translated pages usually keep the English URL behind a language prefix, which keeps them from competing for the Spanish searches your clients type.
I covered how that plays out in search visibility for immigration practices.
When I build a bilingual firm site, a plugin keeps track of which page pairs with which, while a person writes every Spanish page by hand at its own Spanish address.
Settle the ownership question before you start
Not every builder lets you leave with what you made.
Some produce exportable output you own outright, while others keep you inside a publishing environment you can’t leave.
Confirm three things before you spend a weekend on it: that the domain is registered in your firm’s name and you can log into the registrar yourself, that the site files can be exported and moved, and that the analytics history sits in your account rather than a vendor’s.
If any answer is unclear, you are renting rather than owning.
That is the standard I hold myself to on every immigration firm website I build, where the firm owns the domain, the hosting, the files, the analytics and every form submission from day one.
A 12-point checklist before any generated site goes live
If you are doing this yourself, do it properly. Most of it takes an afternoon, and the list works with or without me.
- Submit the contact form from a phone on cell data and confirm the email lands somewhere a human reads.
- Search every live page for leftover placeholder text in brackets, including photo and name markers, plus any lorem ipsum.
- Run the site against WCAG 2.1 AA with a free scanner, then check the three failures scanners catch worst: color contrast, form input labels and missing alt text.
- Navigate the whole site with the keyboard only and confirm you can reach and submit the contact form.
- Confirm every meaningful image has alt text and that each page declares its language.
- Read every page for sentences that promise or imply a result, and delete them.
- Confirm each practice area page describes a matter you handle, and that none describe services you do not offer.
- Verify your bar admissions and the states where you are licensed appear correctly.
- Confirm any testimonial carries whatever disclaimer your state requires.
- Save a dated copy of every page for your advertising archive, and set a recurring reminder on your state’s interval.
- Confirm the domain is in your firm’s name and that the site can be exported.
- Open the site on a real phone and tap the phone number to see whether it dials.
Keep the results, since a dated record of what you checked and when is worth more later than a clean site with no paper trail.
When building it yourself is the right call
If you are solo, the budget is tight, and you need a credible presence while the practice gets moving, generate the site and work the checklist.
That is a reasonable decision, and I would rather you have something honest online than nothing at all.
It stops being the right call once the site is supposed to produce consultations rather than just exist.
That usually surfaces when you start spending on ads and notice the traffic lands somewhere that does not convert, which is the pattern I wrote about after managing $89,000 in Meta spend for one immigration firm.
Traffic is the easy half of the job. The page it lands on and how fast someone answers the inquiry decide whether any of it becomes a signed case.
The short version
AI can build your law firm website, and I use it to build mine.
Whether you should rely on it comes down to whether anyone owns the site after launch and does the work the tool leaves behind.
If you practice in California, ignore anyone selling you accessibility on federal litigation fear, since four of those cases were in the state last year.
Care about it because your Spanish-speaking clients are on the worst-built half of the web, and your intake form is what fails them first; the bar rules haven’t moved, and the archive is your responsibility either way.
If you have the time, work the checklist and keep the money.
If you would rather have the audit, the review, the archive and the Spanish handled by someone who does this for immigration practices, that is what the build covers, and the consultation is a phone call with nothing attached to it.