Your Firm’s Website Structure Is Costing You Matters: The ‘Answer-First’ Architecture for Legal Buyers

Last reviewed by: Lee Thomas, Managing Director, Crescat Digital — [publication date]

Your firm ranks on page one for “employment solicitor Manchester”. The traffic report looks healthy. The enquiry log hasn’t moved in a year, and at the last partners’ meeting someone asked why, and you didn’t have a good answer.

Here is the answer. Buyers can find you. What fails is the few seconds after they arrive, which is a different problem with the same symptom: traffic that looks healthy but doesn’t convert. The page they land on is organised the way your firm is organised, by department, by team, by whoever owns that practice group. They arrived with a question, not an org chart. They skim, they don’t find what they came for, and they go back to the search results or to the next firm on the list their friend gave them.

This article is about the gap between those two things: how your website is built, and what the person reading it needs. It’s aimed at whoever actually owns the website at a UK firm, usually a marketing director or BD lead who owns the site in practice but doesn’t own the pages, because the practice groups do. That distinction turns out to matter more than any of the technical advice, and we’ll get to it.

Fix the mismatch between your structure and your buyers’ questions, and you stop losing matters you were already most of the way to winning.

Table of contents

1. Why your website is organised for your firm, not your buyers
2. What “answer-first” architecture actually means
3. What AI search actually rewards (and what the industry is selling you instead)
4. Why law firm practice area pages fail
5. How should practice area and location pages be structured?
6. How to restructure without fighting partners
7. What to fix first
8. Frequently asked questions

1. Why your website is organised for your firm, not your buyers

Open your firm’s main navigation. It almost certainly mirrors your org chart: practice groups at the top level, teams beneath them, offices hanging off to one side, and somewhere in there, a page about the thing a client actually has a problem with.

This is the natural output of how the site got built. Each practice group signed off its own section, each partner reviewed their own page, and nobody was in the room for the whole thing at once. The site ended up shaped like the firm that made it rather than like the questions it exists to answer, which is the opposite of how structure signals depth to search engines.

Nobody has ever searched for a department. A business owner whose supplier won’t pay an invoice does not search “commercial dispute resolution”. They search something closer to “supplier won’t pay invoice what can I do”. That distance repeats across every practice area you have.

A second reader wants the same thing. When an AI system assembles an answer, it works through the sections of a page rather than reading front to back. It needs the relevant part to be findable and self-contained, and so does a human with a problem and very little patience. You are not building two things. We’ll come back to what that does and doesn’t imply in section three, because that’s where the advice being sold to you goes badly wrong.

The Legal Services Consumer Panel, the statutory body representing the interests of consumers of legal services, asked 3,802 legal services users what would increase their trust in a provider. The top three, in its 2026 tracker survey: specialist expertise (92%), being regulated (91%), and explaining things clearly (91%). Two of those are things your firm either is or isn’t. The third is a thing your website either does or doesn’t do.

Only around one in ten legal clients find their provider by searching the internet. Recommendation and prior use dominate, which is what the Panel’s own data says and what your partners have been telling you for years. But in the Panel’s 2025 survey, 57% of the people who arrived on a personal recommendation still went looking for more information before making contact. And when they look, they frequently don’t find it: the Solicitors Regulation Authority’s most recent evaluation of its transparency rules found 23% of consumers who wanted price information and went to a solicitor’s website couldn’t find it, rising to 31% where the rules specifically applied, against 16% where they didn’t.

So your website is not where clients find you. It’s where they check you out. The referral gets them to the site; the site decides whether they call. That’s a smaller job than marketing budgets are usually justified with, and a more consequential one, because it sits where someone has already decided they might hire you.

↑ BACK TO TOP

2. What “answer-first” architecture actually means

Answer-first means the answer to the question a page is named after appears at the top of that page, in its opening paragraph, before the firm’s history and before the team’s credentials. The rest of the page is downstream of that one decision.

It’s a simple idea that operates at three levels, and most of the advice you’ll read only covers the first.

At page level

The first paragraph of a practice area page answers the question the page exists to answer. If the page is about unfair dismissal, the first thing on it explains what unfair dismissal is and what someone can do about it, in plain language, in the opening paragraph. Everything else follows from there.

At section level

Every heading on the page opens with its own answer, and each section makes sense on its own without the one above it. This is the level Google’s own guidance actually endorses. Its documentation puts it plainly: “People generally appreciate it when web pages are organized by paragraphs and sections, along with headings that provide a clear structure to navigate content.”

At site level

This is the level almost nobody covers, and it costs the most. Answer-first architecture means the pages answering your most valuable questions sit closest to your homepage. Burying your best commercial page three levels down in the navigation is the same fault as burying the answer in paragraph six, executed at the scale of the whole site.

The concept here is click depth: how many clicks it takes to reach a page from your homepage. Google is careful about what this does and doesn’t mean, and it’s worth being equally careful, because the SEO industry is not. Google’s guidance says pages linked directly from the homepage “may be seen as more important, and therefore crawled more often”, but adds, in the same breath, that “this doesn’t mean that these pages will be ranked more highly than other pages on your site”. Separately, it states that while crawling is necessary for a page to appear at all, “it’s not a ranking signal”.

Depth is a cost rather than a penalty. Pages buried deep get found later, get revisited less, and accumulate fewer internal links pointing at them, which is the real mechanism by which they quietly starve. That last part is the fixable one, and it’s worth understanding how internal links pass authority between pages before you start moving things.

While we’re here: the “three clicks from the homepage” rule you’ll hear from a lot of sources has no research behind it. It first appears in a web design book from 2001, which reported it as received wisdom rather than a finding, and it was tested and found wanting in 2003 by a study of 44 users across 620 tasks and over 8,000 clicks, which found no drop-off in success after the third click. The Nielsen Norman Group’s 2019 verdict was blunt: the rule “has not been supported by data in any published studies to date”. Depth is real. The number is made up. Keep the concept and drop the rule, which is a useful habit for this whole subject, as the next section shows.

↑ BACK TO TOP

3. What AI search actually rewards (and what the industry is selling you instead)

If you’ve had an agency in front of you in the last year, you may have been told to break your practice area pages into small chunks (80 to 100 words each, roughly) because that’s the size AI systems prefer to retrieve.

Don’t do it. Google has said so explicitly, and the number you were quoted doesn’t exist.

Start with what’s true, because there’s a real mechanism underneath this. Google runs a system it calls passage ranking, which it describes as “an AI system we use to identify individual sections or ‘passages’ of a web page to better understand how relevant a page is to a search”. Its AI features work by grounding, meaning they retrieve pages from the search index and generate an answer from what those pages contain. So these systems do work through the sections of a page rather than swallowing it whole. Structure matters, which is why how Google reads individual sections of a page has been worth understanding for a while.

Now read that definition again, closely. Passage ranking identifies sections to understand how relevant the page is. It’s a page-ranking signal. It doesn’t score your paragraphs against each other and publish the winner. That gap, between a real mechanism and what it’s claimed to imply, is where a piece of engineering got inflated into a specification nobody ever issued.

The specification is invented twice over. First, the number: the people selling you a chunk length don’t agree with each other. One says 40 to 60 words, another 70 to 150, another 100 to 150, another 100 to 300. One widely-shared piece manages 100 to 150 in its findings and 70 to 150 in its own recommendations, on the same page. The academic research measures in tokens rather than words, reports optimal sizes anywhere between 64 and 1,024, studies systems built over private document collections rather than the open web, and concludes there’s no universal answer anyway.

Second, and decisively: Google has told people to stop. Its guidance on generative AI features carries a section headed “Mythbusting generative AI search: what you don’t need to do”. Chunking is in it:

“‘Chunking’ content: There’s no requirement to break your content into tiny pieces for AI to better understand it. Google systems are able to understand the nuance of multiple topics on a page and show the relevant piece to users. However, sometimes shorter (or longer!) pages can work well depending on your audience and subject matter. There’s no ideal page length, and in the end, make pages for your audience, not just for generative AI search.”

— Google Search Central, Optimizing your website for generative AI features on Google Search, last updated 10 July 2026. Licensed under CC BY 4.0.

Danny Sullivan, Google’s Search Liaison, put it less formally on the company’s own podcast: “turn your content into bite-sized chunks, because LLMs like things that are really bite size… So we don’t want you to do that. I was talking to some engineers about that. We don’t want you to do that. We really don’t.”

The evidence underneath all this is thinner than it looks. The two largest public datasets on what AI systems cite contradict each other. Ahrefs analysed 174,000 pages cited across 560,000 AI Overviews and found essentially no correlation between content length and citation. SE Ranking, publishing the same month, concluded longer content earns more citations. When the two biggest studies in a field disagree that comprehensively, nobody quoting you a precise word count is working from evidence.

So what does work? What Google’s guidance asked for in the first place, and what a decent editor would have asked for regardless: sections that each do one job, with headings that say what’s in them, written so a reader can understand any one of them without having read the one above. That’s answer-first architecture arrived at honestly, not because a machine prefers a word count, but because a section that answers one question completely is useful to a person in a hurry and legible to a system parsing it. Those are the same requirement.

Your practical position: you are being sold a specification that does not exist, by people quoting each other. A firm that rebuilds its practice area pages around an invented number has spent money complying with folklore. A firm that structures its pages so each section answers one question has done what was actually being asked for, and needed no number to do it. For the fuller picture of which signals decide whether AI systems recommend your firm, we’ve written about that separately.

↑ BACK TO TOP

4. Why law firm practice area pages fail

Four patterns account for most of it. You can diagnose all four in about ten minutes.

The department mirror

The page is named after an internal team rather than a problem. “Commercial Dispute Resolution”. “Private Client”. “Contentious Probate”. These are the names on the door of the practice group, and no client has ever typed one of them into a search box. The page ranks for nothing because it’s answering a question nobody asked.

The credentials-first page

The page does answer the buyer’s question, eventually, after 200 words about the firm’s founding, the team’s accolades, and a sentence about being “committed to excellence”. A human skimmer bounces before reaching the answer, and so does anything reading the page for a useful section. The information is all there, behind the wrong things.

The partner’s page

Written by, and effectively owned by, a senior fee earner who cares about it. It reads like a conference paper: technically excellent, thorough, and commercially inert. It is also, as far as your ability to change it goes, made of granite. It’s the pattern most likely to be on your site and least likely to be in anyone else’s advice, because it’s a political problem wearing a technical costume. We’ll deal with it in section six.

The orphaned page

A genuinely good page that nothing links to. It sits four or five clicks from your homepage, gets found late, gets revisited rarely, and receives almost no internal signal that it matters. In our audit work it is very often the single highest-value commercial page on the site, the one that would earn real matters if anyone could reach it. Worth checking how Google actually crawls your site to find yours.

Then there’s duplication. Most firms with more than one office have several nearly-identical location pages competing for closely-related searches, splitting the signal so none of them wins.

↑ BACK TO TOP

5. How should practice area and location pages be structured?

The model is hub-and-spoke: one broad page per practice area, with a spoke page for each specific problem beneath it. The hub answers the general question and links down to the spokes; each spoke links back up to the hub. This concentrates the topical authority that tells search engines your firm knows a subject deeply rather than glancingly, and it distributes authority from your strongest pages to the ones that need it. We didn’t invent it, and how internal and inbound links work together is the mechanism underneath.

What isn’t well understood is what each type of page has to do. This is where most firms go wrong: they build the right shape and fill it with the wrong content.

Location pages that aren’t duplicates

Two different problems live here, and they need different answers.

Building new ones. A location page is worth building when it’s genuinely different: local courts, the local tribunal, how the process works there, the named people in that office. That’s substance a reader would notice. Swapping “Leeds” for “Bradford” in an H1 and changing nothing else is the same page twice.

Dealing with the ones you already have. More common, and more uncomfortable, because the answer is usually to have fewer pages. Consolidating three weak location pages into one strong one is a gain rather than a loss, though the instinct that deleting pages is dangerous is real and worth addressing head-on.

This is one of the few places in the subject where Google publishes a rule rather than folklore. Its spam policies list, under “doorway abuse”, “[h]aving multiple domain names or pages targeted at specific regions or cities that funnel users to one page”, and “[c]reating substantially similar pages that are closer to search results than a clearly defined, browseable hierarchy”.

Read that first item again with your own location pages in mind. Six near-identical town pages for places where you have no office, all funnelling to one central contact form, is close to a literal description of it.

There’s no bright line, and anyone offering you a threshold (a number of pages, a similarity percentage) is inventing it. What Google gives you is criteria. Three questions, drawn from its own wording:

• Is the page the destination, or does it funnel people elsewhere?
• Does it add real value, or is it substantially similar to its siblings?
• Can you reach it by browsing the site, or does it exist as an island only reachable from search?

A town page with a staffed office and genuinely local content fails none of those. Six duplicates for towns you’ve never had a desk in fail most.

If you do consolidate, the mechanics matter, and Google’s guidance is specific about which tool does which job.

And before you merge anything, check the page isn’t quietly earning traffic from something you weren’t thinking about. How to diagnose overlapping pages in Search Console walks through it.

One note on vocabulary, because it shapes how you think about the problem. The industry calls this cannibalisation: the idea that two of your own pages compete for the same search and cancel each other out. The term does more harm than good. Google has never used it, doesn’t document it, and John Mueller has pushed back on how the concept is commonly applied, noting that three pages appearing in the same result “doesn’t seem problematic to me just because it’s ‘more than 1’”. What’s real is what sits underneath the label: near-duplicate pages competing, thin pages, unfocused pages, pages nobody links to. Four different problems with four different fixes, rolled into one word that makes them harder to solve rather than easier. Diagnose the fault, not the label.

↑ BACK TO TOP

6. How to restructure without fighting partners

Everything above assumes you can change your website. You probably can’t, or at least not in the way the advice implies.

You own the site. You don’t own the pages. The practice groups do, in every way that matters. They wrote them, they approve them, and they will hear “we’re restructuring your section” as “we’re taking your section away”. A structurally perfect plan that requires eleven partners to give up territory at once will not survive its first meeting. Worse, proposing it burns the credibility you need to propose anything smaller afterwards.

So sequence it.

1. Start where nobody owns the page. Location pages, service pages that were nobody’s project, orphaned content, the things that got built once and forgotten. Political cost is near zero, and this is frequently where the biggest structural gains are anyway.
2. Measure what changes. Not traffic. Enquiries from those specific pages, before and after, over a defined window. You are building evidence, and traffic is not evidence to this audience.
3. Bring the evidence, not the architecture. No partner will engage with a site map. They will engage with: “the three pages we restructured are now producing four enquiries a month between them. Yours isn’t in that comparison.” Let them ask the next question.
4. Then ask for the territory. Once the pattern is demonstrated on pages nobody was defending, the conversation about the litigation team’s page is a completely different conversation, because you’re no longer asking them to trust a theory.

This is slower than doing it properly in one pass. That’s the trade, and it’s worth making, because the one-pass version mostly doesn’t happen. The firms that fix their structure are the ones whose plan survived contact with the partnership, not the ones with the best plan.

If this pattern feels familiar, it’s the same problem we wrote about when examining the SEO risks contained in a firm’s news and insights section: different territory, identical politics.

↑ BACK TO TOP

7. What to fix first

Don’t rebuild the website.

That’s the instinct when the structural argument lands, and it’s the wrong one. A rebuild is expensive, slow, and mostly solves a different problem than the one you have. We’d rather tell you that than sell you one.

Almost all of the gain is in reordering what you already have. Move the answers up. Consolidate the duplicates. Link your hubs to your spokes. Pull your commercial pages closer to the homepage. None of that requires a new website, a new CMS, or a six-figure project. It requires knowing which pages matter and being allowed to touch them, and the second half of that is the hard part.

Come back to where this started. Your website isn’t where your clients find you; it’s where they check you out, and most of them do. Somebody your best client trusts told them to call you, and before they did, they went and looked. What they found was your org chart. The real question is whether the person who’s already decided to hire you can confirm they’re right before their patience runs out.

Fix that, and the enquiry log moves, not because you’ve been found by more people, but because you’ve stopped losing the ones who’d already found you.

↑ BACK TO TOP

8. Frequently asked questions

How deep should pages be on a law firm website?

Shallower is better, but there’s no magic number, and the widely-quoted “three-click rule” has no research behind it. Google says pages linked from the homepage may be crawled more often, while explicitly adding this doesn’t mean they rank more highly. Depth is a discovery and linking cost, not a penalty. If a page that should be earning matters is buried, that’s worth fixing whatever the click count says.

Do law firms actually need separate location pages?

Only where the firm genuinely serves the location and can say something specific about it: the local court, the local process, the people in that office. If you could swap the town name and change nothing else, you don’t have a location page, you have a duplicate. Sometimes the right answer is fewer pages, not more. Google’s spam policies specifically flag pages targeted at cities that funnel users to one page.

Should we rebuild the website or restructure what we have?

Restructure, almost always. Most of the value is in reordering what already exists: moving answers to the top of pages, consolidating duplicates, linking hubs to sub-pages, and pulling commercial pages closer to the homepage. A rebuild is expensive, slow, and tends to reproduce the same structural fault in a nicer template, because the politics that created the structure don’t change.

How long does a restructure take to show results?

It varies by practice area and how competitive your area is, so treat anyone quoting you a fixed timeline with suspicion. Structural changes to pages that already rank tend to show up faster than work on pages starting from nothing. We’ve set out realistic timelines by practice area separately.

Should we restructure our pages for AI search?

Structure your pages for readers, which is exactly what Google’s own guidance asks for, and ignore anyone selling you a word count. Google’s documentation is explicit: “There’s no requirement to break your content into tiny pieces for AI to better understand it… There’s no ideal page length.” Sections that each answer one question completely serve both readers and AI systems. There’s no separate AI version of this job.

↑ BACK TO TOP

Sources

• Legal Services Consumer Panel, 2026 Tracker Survey: How consumers are using legal services, published 8 July 2026. Based on a survey of 3,802 legal services users in England and Wales, carried out between 23 February and 13 April 2026 by M·E·L Research.
• Legal Services Consumer Panel, 2026 Tracker Survey: How consumers are choosing legal services, published 8 July 2026.
• Legal Services Consumer Panel, 2025 Tracker Survey: How consumers are choosing legal services, published 7 July 2025. Based on a survey of 3,750 legal services users.
• Solicitors Regulation Authority, Year Three Evaluation of the SRA Transparency Rules, 24 October 2023.
• Google Search Central, Optimizing your website for generative AI features on Google Search, last updated 10 July 2026.
• Google Search Central, A guide to Google Search ranking systems, last updated 10 December 2025.
• Google Search Central, Myths and facts about crawling, last updated 18 December 2025.
• Google Search Central, Spam policies for Google web search, last updated 15 May 2026.
• Google Search Central, How to specify a canonical URL with rel=“canonical” and other methods, last updated 10 July 2026.
• Danny Sullivan, “SEO, AIO, GEO, your site, & third-party support to optimize for LLMs”, Search Off the Record podcast, Google, January 2026.
• John Mueller, Google, on keyword cannibalisation, 20 September 2025.
• Ahrefs, “Short vs. Long Content in AI Overviews: The Data Says Both Work”, Despina Gavoyannis and Xibeijia Guan, 3 December 2025.
• SE Ranking, “How to optimize for AI Mode”, Yulia Deda, 15 December 2025.
• Page Laubheimer, “The 3-Click Rule for Navigation Is False”, Nielsen Norman Group, 11 August 2019.
• Joshua Porter, “Testing the Three-Click Rule”, User Interface Engineering, 2003.