AI Tools for Lawyers – SEO Improvements

The H1 & Meta Tags — title rebuilt on a KD 22 term · three descriptions rewritten · both social images were EMPTY · og:url was relative

The old descriptions promised three things the product does not do — document drafting, smart document generation and summarising legal research. That is why five keyword rows in the table below show a negative delta: they were removed on purpose, not lost. The H1 «AI Tools for Lawyers» (590 vol / KD 55 / $18.39) was left unchanged; the nav and footer use shortname="AI Tools" via GetShortName, so changing it would not have broken them — it was simply the best AI-qualified term available.

Content/ai-tools/En/articlemetas.cshtml · App_Data/ServiceEntities.xml · WAS = git HEAD before 2026-09-08

@* TITLE — REBUILT ON THE DATA, NOT ON THE AI CLUSTER. The obvious candidate "AI Tools for Law Firms" is 170 vol at KD 61, and "legal ai tools for law firms" is 10 vol at KD 52 — the AI-qualified long forms have almost no volume. The title now leads with "legal workflow automation": 480 vol, KD 22, CPC $52.22, and already present on the page. *@
@* was: AI Tools | {CenterName} — four words, no qualifier, no legal term *@
<title>AI Tools for Lawyers - Legal Workflow Automation | {CenterName}</title>

@* DESCRIPTION — the old one promised document drafting and legal-research summarisation. THE PRODUCT DOES NEITHER. Rewritten around what the tools actually do, carrying "AI matter management", "legal contract review" and "legal document review". *@
@* was: Supercharge your legal practice with Maatdesk AI. Automate document drafting, summarize complex legal research, and enhance decision-making... *@
<meta name="description" content="AI tools for lawyers that read your firm's own files: AI matter management, legal contract review and legal document review, inside the case file." />

@* OG:TITLE — "AI Tools & Automation" is 20 vol. Replaced with the two feature terms that carry real volume at low difficulty: matter management (590 / KD 15) and contract review (1,300 / KD 29). *@
<meta property="og:title" content="AI Tools for Lawyers - AI Matter Management & Contract Review | {CenterName}" />

@* OG:DESCRIPTION — was generic "leverage artificial intelligence" copy built on a KD 100 non-legal term. Now describes the three reports the product actually returns. *@
<meta property="og:description" content="Legal AI tools for law firms. Maatdesk summarizes a matter, scores a contract and checks a pleading before filing..." />

@* OG:URL — WAS RELATIVE ("/ai-tools"). Open Graph requires an absolute URL; a relative one is silently dropped by most consumers. *@
<meta property="og:url" content="https://maatdesk.com/ai-tools" />

@* SOCIAL IMAGES — og:image AND twitter:image were BOTH EMPTY, and twitter:card was "summary". Cheapest fix on the page: the asset already existed on disk. width/height/alt added so the card renders without a fetch. *@
<meta property="og:image" content="themes/maatdesk/images/services/thumbnails/ai-tools.jpg" />
<meta name="twitter:card" content="summary_large_image" />

@* TWITTER META — THE WEAKEST BLOCK ON THE PAGE. Both of its distinctive phrases were dead: "automated summaries" (no reportable volume, KD 69) and "smart document generation" (0 volume, KD 22). Neither survived. *@
@* was title: AI Legal Tools & Automation  |  was desc: ...smart document generation, Maatdesk is the future of legal tech. *@
<meta name="twitter:title" content="Legal AI Tools for Law Firms" />
<meta name="twitter:description" content="Legal workflow automation for law firms: AI matter management, contract analysis and legal document review..." />

@* H1 — UNCHANGED. "AI Tools for Lawyers", from ServiceEntities.xml id="ai-tools" name attribute. 590 vol / KD 55 / $18.39: the best AI-qualified term in the data, and the page had it in the meta description ONLY. *@
@* INTRO PARAGRAPH — WAS EMPTY. The briefintro attribute now carries 78 words, rendered under the H1 by ServicePageSkin.cshtml. *@

@* UNCHANGED AND CORRECT: canonical https://maatdesk.com/ai-tools · article:published_time 2024-09-11. *@

The title rewrite reversed an earlier recommendation, and the reversal is the finding. The first draft of this analysis proposed «AI Tools for Law Firms». The bulk pull killed it: 170 vol at KD 61, against «legal workflow automation» at 480 vol / KD 22 / $52.22 — a term already sitting on the page, in a Twitter meta and one FAQ answer, with no heading, no title and no body copy behind it.
The pattern that decided every meta on this page: AI-qualified terms are high-difficulty and low-CPC; «legal X software» terms are low-difficulty and high-CPC. Every AI query in the GSC data that Semrush could price came back at $0.00 CPC.

Keyword Coverage — Was / Now — 136 rows · 40 rose, 26 fell, 70 unchanged · occurrences 128 → 390 · keywords present 57 → 56

Keyword breadth did NOT go up — depth did. 57 of the 136 rows were present before, 56 after. The page did not add vocabulary; it removed three false product claims and 26 generic filler phrases, and put weight behind the terms that were already true. A Δ in this table counts a string. It is not a ranking signal, and higher is not automatically better.

SR-GAP 753 kw · SR-BULK1 45 kw + SR-BULK2 36 kw, Semrush US 2026-09-08 · GSC 2026-06-07 → 2026-09-06 · WAS = git HEAD, NOW = working tree 2026-09-09

Keyword	Cat	Vol	KD	CPC	AIO	GSC impr	Pos	Was	Now	Δ	Note
legal workflow automation	PRIM	480	22	52.22	Y	S: 12	6.5	0	9	+9	⭐ THE BEST TERM THE PAGE ALREADY OWNED. KD 22 at $52.22 CPC, and it lived in a Twitter meta and one FAQ answer. Now in the title
matter management	FEAT	590	15	17.12	Y	P: 7	30.0	0	14	+14	⭐ LOWEST KD OF ANY TERM ON THE PAGE. «matter management software ai» is the one on-target non-brand query the page surfaces for
contract review	FEAT	1300	29	19.43	Y	S: 140	55.6	1	20	+19	⭐ 43× the volume of «Contract Reviewer», the H3 the page actually used (30 vol / KD 55)
legal document review	FEAT	480	21	10.02	Y	0	—	0	14	+14	⭐ The winnable term the «Pleading Reviewer» card should have carried all along — pleading* is 0 kw / 0 vol in the whole 753-row gap set
legal AI	PRIM	12100	79	9.02	—	S: 315	~30	1	18	+17	darrow.ai ranks #7. Was absent from the title entirely
AI tools for lawyers	PRIM	590	55	18.39	—	0	—	2	13	+11	The H1 term. GSC-KW confirms ZERO impressions in 92 days, and it had existed in the meta description only
AI tools	PRIM	27100	91	3.89	Y	0	—	12	27	+15	⚠ THE PAGE'S MOST-REPEATED PHRASE IS ITS WORST. KD 91, $3.89, and not even a legal term. It rose with the word count — it is not a target
AI for law firms	PRIM	880	54	10.45	—	S: 46	34.7	0	4	+4	H2 + schema audience + VideoObject
AI legal tools	PRIM	320	66	24.89	—	0	—	3	2	−1	Commercial intent, high CPC, but KD 66 — not worth a heading
artificial intelligence	PRIM	1830000	100	3.53	Y	0	—	1	3	+2	KD 100, generic, non-legal. Ignore as a target
legal practice management software	SUPP	4400	30	41.03	Y	S: 434	7.6	1	1	=	⭐ THE BIGGEST SINGLE MISS. 4,400 vol at KD 30 and $41.03, buried in a JSON-LD `applicationSubCategory` attribute with no textual support anywhere
legal document management	FEAT	1600	34	34.03	Y	S: 106	2.2	1	1	=	Schema-only. KD 34 at $34.03, already ranking sitewide at position 2.2
case management	FEAT	8100	27	4.67	Y	S: 414	58.6	1	3	+2	KD 27 at 8,100 vol but $4.67 CPC. The commercial form is «legal case management software» below
contract analysis	FEAT	480	27	13.09	Y	0	—	0	5	+5	KD 27, and it had been named in the video schema only, nowhere visible
legal contract review	FEAT	140	29	15.49	Y	0	—	0	5	+5	Qualified variant, same low KD
Matter Summarizer	FEAT	n/v	82	—	—	0	—	1	9	+8	⚠ Worst H3 on the page: no reportable volume AND KD 82. Kept as branding; «matter management» now carries the term
Contract Reviewer	FEAT	30	55	15.32	Y	S: 140	55.6	1	8	+7	The H3 as written = 30 vol. Kept as a product name, subhead carries «contract review»
Pleading Reviewer	FEAT	n/v	20	—	—	0	—	1	8	+7	KD 20 but no volume; 0 pleading keywords in 753 gap rows, 0 GSC queries
document drafting	FEAT	70	30	6.55	Y	S: 139	7.4	1	0	−1	⚠ REMOVED AS A FALSE CLAIM — the product does not draft documents. Not a loss
legal research summarization	FEAT	n/v	24	—	—	0	—	0	0	=	⚠ Also a false claim, and no volume as phrased. The real term is «ai for legal research» — deliberately left at zero
automated summaries	FEAT	n/v	69	—	—	0	—	1	0	−1	No volume + KD 69. Dropped from the Twitter meta
smart document generation	FEAT	0	22	0.00	—	0	—	1	0	−1	KD 22 but 0 volume. A dead phrase, and a claim the product could not support
court filing	FEAT	590	51	0.96	Y	S: 48	61.8	1	0	−1	Volume exists, but $0.96 CPC — low commercial value
decision support	FEAT	210	41	0.00	Y	0	—	1	0	−1	$0.00 CPC — no commercial intent behind it
law firms	SUPP	6600	75	0.00	Y	S: 8	35.5	3	14	+11	KD 75 and $0.00 CPC — a qualifier, not a target on its own
attorneys	SUPP	9900	77	14.08	—	S: 9	33.4	4	15	+11	Local-pack SERP dominated by directories. Not winnable for a SaaS page
legal tech	SUPP	1600	42	7.51	Y	0	—	1	0	−1	Went out with the Twitter meta rewrite
professional judgment	SUPP	170	16	0.00	Y	0	—	2	2	=	KD 16 but $0 CPC and non-legal intent. Kept for E-E-A-T, not for ranking
legal case management software	XSELL-Q	2400	35	43.06	Y	S: 414	58.6	0	2	+2	⭐ Highest-volume winnable commercial term in the whole dataset. Belongs to its own landing page, not this one
legal contact management software	XSELL-Q	140	10	89.11	Y	S: 227	26.2	0	0	=	⭐ LOWEST KD AND HIGHEST CPC IN THE ENTIRE DATASET. Its own page owns it — correctly left at zero here
legal billing software	XSELL-Q	1900	38	75.63	Y	S: 56	44.3	0	0	=	$75.63 CPC at KD 38 — second-highest CPC in the dataset. Owned by /legal-billing-software
legal task management software	XSELL-Q	260	24	50.74	Y	S: 434	31.2	0	0	=	434 impressions stuck at position 31
legal calendaring software	XSELL-Q	480	21	36.41	Y	S: 374	32.0	0	0	=	374 impressions stuck at position 32
law firm client portal	XSELL-Q	210	18	15.96	Y	S: 155	26.3	0	0	=	The one query family that has ever produced a click sitewide
law firm mobile app	XSELL-Q	90	15	13.47	Y	S: 0	—	0	0	=	KD 15 vs KD 79 for the generic «mobile app» — same product, one fifth the difficulty
legal ai tools	GAP	1900	65	25.30	—	0	—	0	9	+9	darrow.ai #2, legora #1. High CPC, high KD
legal artificial intelligence tools	GAP	720	49	25.30	—	0	—	0	3	+3	⭐ Best volume/KD ratio in the AI cluster, and no competitor in the top 15
ai lawyer	GAP	9900	61	14.62	—	0	—	0	2	+2	Second-biggest term in the gap set. Appears only inside FAQ 9, as a denial
ai for legal research	GAP	3600	54	17.57	—	0	—	0	0	=	⚠ DELIBERATELY LEFT AT ZERO — the product does not do legal research. research* is 43 kw / 9,500 vol and none of it is honestly ours
lawyer ai	GAP	3600	61	9.02	—	0	—	0	0	=	NOT FORCED — no sentence places that word order naturally
ai tools for law firms	GAP	170	61	17.14	Y	S: 6	35.2	0	5	+5	The rejected title candidate: only 170 vol at KD 61 — weaker than it looks
legal contract ai tools	GAP	30	63	100.09	—	0	—	0	0	=	Highest CPC in the 753-keyword gap file at $100.09 — and only 30 vol
legal ai billing pricing	GSC-Q	n/v	31	—	—	S: 7	2.43	0	0	=	⭐ Position 2.43 with 0 clicks. The highest-intent unserved query in the dataset
ai enabled practice management pricing	GSC-Q	n/v	3	—	—	S: 7	3.0	0	0	=	⭐ KD 3, Transactional intent, position 3, zero clicks
how much does legal ai software cost	GSC-Q	n/v	25	—	—	S: 20	4.8	0	0	=	Page 1, 0 clicks — and no visible pricing copy on the page at all
custom legal ai implementation cost	GSC-Q	n/v	10	—	—	S: 14	7.5	0	0	=	KD 10, page 1, 0 clicks. Direct FAQ candidate
legal ai clause library migration	GSC-Q	n/v	7	—	—	S: 1	7.0	0	0	=	⭐ KD 7 — the lowest difficulty of any query in the dataset. Untapped FAQ topic
migrate matters/contacts/documents to legal ai	GSC-Q	n/v	11	—	—	S: 4	3.0	0	0	=	KD 11 at position 3. The direct replacement for the weakest original FAQ
role based access legal ai	GSC-Q	n/v	29	—	—	S: 1	5.0	0	0	=	KD 29 at position 5, and the page said «permissions» exactly once and never expanded
matter management software ai	GSC-Q	10	16	0.00	—	P: 7	30.0	0	0	=	⭐ THE ONE ON-TARGET NON-BRAND QUERY THE PAGE SURFACES FOR
ai for legal software	GSC-Q	0	67	0.00	Y	S: 44	45.5	0	0	=	Largest sitewide AI query by impressions — yet Semrush reports 0 volume for it
Monthly Plan ($39/user/mo)	COMM	n/d	—	—	—	S: 7	2.4	5	19	+14	Offer schema. Ranks at position 2.43 with 0 clicks and no visible price on the page
maatdesk login	GSC-Q	50	0	0.00	—	P: 267	6.81	0	0	=	⚠ 267 of the page's 302 de-anonymised impressions. /ai-tools should not be the login result
Maatdesk (brand)	ENT	170	17	2.48	—	P: 291	3.2	23	69	+46	Sitewide: 1,239 impr / 235 clicks / position 1.05. Brand is 97% of this page's de-anonymised impressions

Full table: 136 rows, merging all on-page keywords, the 753-keyword gap set and every live GSC query. The 53 rows above are the decision-relevant ones. Totals: occurrences 128 → 390 · 40 rows rose, 26 fell, 70 unchanged · keywords present 57 → 56.
Scope, counted identically on both sides — and it differs from the home-page report: rendered text only, meaning H1 + briefintro from ServiceEntities.xml + metas + body copy AND the JSON-LD. Razor comments excluded. Matching is case-insensitive substring; a row whose keyword lists alternatives with «/» sums each alternative.
The 26 falls are the point, not a regression. Five are removed product claims (document drafting, smart document generation, automated summaries, legal research summarisation, decision support). Ten are the unqualified cross-sell generics that left with the link block — correct, because those terms belong to the pages that own them. The rest are generic body words (insights, patterns, accuracy, efficiency) that were never targets.
WHERE THE TOOLS DISAGREE: «ai for legal software» carries 44 real GSC impressions and 0 Semrush volume; «legal ai billing pricing» has no reportable volume and ranks at position 2.43. Every n/v row in the GSC-Q block is real demand the keyword tool cannot see.

What the Data Says — 753 gap keywords + 81 bulk keywords + the GSC query set, and the one pattern that decided every edit

GSC 2026-06-07 → 2026-09-06: /ai-tools 750 impressions · 0 clicks · 0.00% CTR · avg position 6.22 · sitewide 15,509 impr / 241 clicks

1. The AI vocabulary is the wrong vocabulary THIS FINDING DROVE EVERY OTHER DECISION

AI cluster	Vol	KD	CPC	«legal X software» cluster	Vol	KD	CPC
legal ai	12,100	79	$9.02	legal contact management software	140	10	$89.11
ai lawyer	9,900	61	$14.62	legal billing software	1,900	38	$75.63
ai for legal research	3,600	54	$17.57	legal task management software	260	24	$50.74
legal ai tools	1,900	65	$25.30	legal case management software	2,400	35	$43.06
ai tools for law firms	170	61	$17.14	legal practice management software	4,400	30	$41.03
legal ai tools for law firms	10	52	$0.00	legal calendaring software	480	21	$36.41

AI terms are high-difficulty and low-CPC. Software terms are low-difficulty and high-CPC. And every AI query in the GSC data that Semrush could price came back at $0.00 CPC. That is the whole argument for building the title on «legal workflow automation» rather than on any AI phrase — and for not chasing «ai lawyer» (9,900 vol) at all.

2. The page targeted the hardest term and ignored the easy ones it already contained

«AI tools» — KD 91, $3.89 CPC, not a legal term — was the most-repeated phrase on the page. Every term below was already present in the copy and targeted by nothing: no title, no heading, no meta. matter management 590 / KD 15 · legal document review 480 / KD 21 · legal workflow automation 480 / KD 22 / $52.22 · contract review 1,300 / KD 29 · legal practice management software 4,400 / KD 30 / $41.03 · legal document management 1,600 / KD 34 / $34.03.
Combined that is 9,270 monthly volume at KD 15–34, sitting under a phrase worth $3.89 at KD 91.

3. The three feature H3s were product names with no demand SUBHEADS NOW CARRY THE TERM

H3 as written	Vol	KD	The keyword sitting underneath it	Vol	KD
Matter Summarizer	n/v	82	matter management	590	15
Contract Reviewer	30	55	contract review	1,300	29
Pleading Reviewer	n/v	20	legal document review	480	21

The product names were KEPT as branding — they name real features and the screenshots use them. What changed is that the H3 or its subhead now carries the real term. «Pleading Reviewer» is the sharpest case: pleading* returns 0 keywords and 0 volume across all 753 gap rows, while «legal document review» is 480 vol at KD 21.

4. Pricing intent is ranking, unserved, and trivially winnable STILL OPEN — NO VISIBLE PRICING COPY

Query	KD	Position	Intent
ai enabled practice management pricing	3	3.0	Transactional
custom legal ai implementation cost	10	7.5	Informational
how much does legal ai software cost	25	4.8	Informational
legal ai billing pricing	31	2.43	Transactional

Four queries at KD 3–31, all ranking on page 1, all at zero clicks. The three Offers — $0 six-month trial, $39/user/mo, $29/user/mo billed yearly — exist only in the Offer schema. There is no visible pricing copy on the page. This is the cheapest remaining conversion opportunity in the whole dataset and it is not yet done.

5. The old FAQ block answered questions nobody asks — four real ones already rank

All five original FAQ questions were brand-scoped with no reportable volume, and one carried KD 66 with no volume at all. Meanwhile these four were already ranking at positions 3–7 with nothing on the page to serve them: legal ai clause library migration — KD 7, position 7.0 · migrate matters/contacts/documents to legal ai — KD 11, position 3.0 · custom legal ai implementation cost — KD 10, position 7.5 · role based access legal ai — KD 29, position 5.0.
These are not keywords to chase; they are queries Google is already showing the site for and getting no clicks on.

6. Cross-sell terms in the featureList were unqualified generics ALL 13 REPLACED WITH 6 AI CAPABILITIES

Generic in featureList	KD	Qualified form	KD	CPC
Time Tracking	100	legal time tracking software	30	$35.65
Mobile App	79	law firm mobile app	15	$13.47
Task Management	77	legal task management software	24	$50.74
Contact Management	49	legal contact management software	10	$89.11
Billing	50	legal billing software	38	$75.63
Client Portal	40	law firm client portal	18	$15.96
Calendaring	26	legal calendaring software	21	$36.41
E-Signature	70	legal e-signature software — THE EXCEPTION	64	$13.22

Every qualified form except e-signature is dramatically easier than its generic — and «legal e-signature software» is 10 vol at KD 64, the only qualified form worse than the generic it replaces.
THE ACTION TAKEN WAS NOT TO ADD THE QUALIFIED FORMS HERE. The featureList was rewritten to list the 6 AI capabilities instead of the platform's 13 modules, and the link block that carried these terms was removed. Those keywords dropped to 0 on this page on purpose: the pages that own them should rank for them.

7. Semrush family totals in the 753-keyword gap set — where the topic's mass actually sits

tool* — 105 kw / 11,460 vol
research* — 43 kw / 9,500 vol NOT OURS
software* — 49 kw / 5,600 vol
attorney* — 22 kw / 4,610 vol
platform* — 24 kw / 3,430 vol
assistant* — 20 kw / 3,340 vol UNTOUCHED POSITIONING
document* — 23 kw / 1,770 vol
draft* — 25 kw / 1,580 vol NOT OURS
case* — 13 kw / 800 vol
matter* — 3 kw / 790 vol
automat* — 7 kw / 530 vol
review* — 4 kw / 270 vol
contract* — 2 kw / 240 vol
workflow* — 2 kw / 120 vol
summar* — 1 kw / 40 vol
pleading* — 0 kw / 0 vol AN H3 SITS ON THIS

The two largest families after «tool» are research and drafting — 68 keywords and 11,080 monthly volume that this product cannot honestly claim. The families the product does serve (matter, contract, review, workflow, summar) total 12 keywords and 1,460 volume. That is the real ceiling of this page as an AI-topic page, and it is why the recommendation is to anchor on the software vocabulary instead.

8. Traffic mix — 97% brand, and one query is two thirds of it

Only 7 of the page's queries were de-anonymised, covering 302 of 750 impressions — and 5 of the 7 are brand navigational. «maatdesk login» alone is 267 of those 302, at position 6.81. Exactly 8 impressions came from non-brand AI queries. Devices: Desktop 254 impressions at position 6.8 · Mobile 48 at 8.19. Countries: US 293 of 302 at position 6.39, then TH 4, HK 2, BR 1, VN 1, SG 1.
ZERO CLICKS ON 750 IMPRESSIONS at average position 6.22, and AI Overviews appear on 40+ of the target SERPs, which compresses organic CTR. But the page also had no social image, a «summary» card and a four-word title — fix the snippet before chasing new rankings.
«maatdesk login» is a routing problem, not an SEO one: /ai-tools should not be the result Google shows for it.

Content Changes — 255 → 3,253 body words across three passes · 11 screenshots converted to HTML · 5 new sections

Every change below is also documented in a Razor comment directly above the markup it affects, with the help-centre or Semrush source it came from. Nothing here is a new product claim — the three report panels reproduce real output, the client-data section carries two external citations, and the three removed capability claims were removed precisely because nothing substantiated them.

Content/ai-tools/En/article.cshtml · ServicePageSkin.cshtml · App_Data/ServiceEntities.xml · 2026-09-08 and 2026-09-09

The single biggest indexability fix — 11 screenshots holding ~490 words no crawler could read RESOLVED 09-08

Eleven .webp screenshots (ms0–ms4, cr1–cr3, pr0–pr2) carried roughly 490 words of report text as pixels, and none of them had an alt attribute. All 11 were rebuilt as three .ai-report HTML panels reproducing the real report output — the text is now indexable, selectable and readable by assistive technology, and the AI-report CSS was added to both custom.css and main.min.css.
THE PANELS WERE THEN AUDITED AGAINST THE SOURCE IMAGES ON A SECOND PASS, and five errors were found in the first transcription: the third Contract Review box was labelled «Jurisdiction» where cr1 says «Compatibility» and also carries the agreement type · «Missing or Weak Clauses» listed Confidentiality, which appears nowhere in the source — cr2 names SLA, data security and indemnification, and describes them as absent, not one-sided · the Pleading Review line «all mandatory procedural elements are present» was inferred from the 85/100 score and is in no source image · «et al.» was missing from the defendant · and the scoring criteria read «clarity, fairness and compliance» in four places where cr1 says «clarity, balance, risks, and compliance».
NOTE FOR FUTURE EDITORS: pr0/pr1 and pr2/pr3 are two DIFFERENT runs. pr2/pr3 analyse a charitable pledge agreement and their own output says «not a legal pleading», so they are deliberately not represented. The panel reproduces the Notice of Appearance run only.

Sections added — and one that was built and then deleted

The definitional section, then reframed. Written at category level first; it measured 1.1 brand references per 100 words, the lowest on the page, and was rewritten to product level — «In Maatdesk, legal AI tools are three specific things…» MEASURED, THEN REWRITTEN
«Every Review Returns the Same Sections», then cut by 46%. Its two bulleted lists were reduced to section names only: 14 of the 23 labels were already shown with real values in the sample panels above them, so the definitions were duplication. 351 → 191 words with zero keyword loss 160 WORDS CUT
«What Maatdesk AI for Lawyers Actually Changes» — three moments built entirely from findings already on the page: the 35/100 contract, the pleading with no judge assigned. No new claims NEW
«What Your Firm Gains» — 12 items, split attorney-level and firm-level NEW
«How Maatdesk AI for Law Firms Handles Your Client Data», carrying both external citations. Every claim sourced; nothing asserted about where document text travels or what is retained, because that is documented nowhere NEW · 2 CITATIONS
The six feature cards rewritten, three added — Saved AI Responses, Firm-Level AI Integrations, Bring Your Own AI Provider 6 → 9 CARDS
«On This Page» index (.toc-card) under the video, 8 entries, with scroll-margin-top: 110px on all 8 heading ids for the sticky header NEW
BUILT AND THEN REMOVED: «Works with the rest of Maatdesk». It put 16 links in one paragraph, 8 of them exact-match anchors. Replaced with 4 contextual links inside sentences that already named those products — outbound links 18 → 9 DELETED ON PURPOSE

Headings — 3 H2s became 8, and 4 of the originals were reworded to name the brand

#	H2 now	Was
1	What Are Maatdesk's Legal AI Tools for Law Firms?	Only three H2s existed: «Legal AI Tools That Support Your Practice - Not Replace It» «Advanced AI Tools for Modern Law Firms» «FAQs»
2	Legal Workflow Automation Built Into Maatdesk — the KD 22 / $52.22 term, in a heading for the first time
3	Reports Your Attorneys Review and Approve — Not AI That Replaces Them
4	Every Maatdesk Review Returns the Same Sections	new
5	What Maatdesk AI for Lawyers Actually Changes	new
6	How Maatdesk AI for Law Firms Handles Your Client Data	new
7	What Your Firm Gains from Maatdesk AI Tools for Lawyers	new
8	Frequently Asked Questions About Maatdesk Legal AI Tools	«FAQs»

Four of the original H2s were reworded to name Maatdesk, varying the brand position so the «On This Page» index does not stutter. No keyword was lost in the rewording.
The H2 that was dropped is worth recording: «AI that supports your practice» was the page's top H2 and it targeted nothing — a slogan with no keyword behind it at all.

US spelling pass SITE TARGETS THE US

summarise / summarised / summarises · organisations · totalling · standardised · licence → summarize / summarized / summarizes · organizations · totaling · standardized · license Semrush data is pulled US, the schema declares inLanguage en-US, and the GSC data is 293 of 302 impressions from the United States. The copy had been drifting British.

FAQ Block — 5 → 9 — the original five were brand-scoped with no volume · four real questions rank at positions 3–7 and are still unserved

Every one of the five original questions was brand-scoped and had no reportable Semrush volume — and one of them combined KD 66 with no volume at all, the worst pairing anywhere in the dataset. The block grew to 9 questions with the accordion and the FAQPage schema kept in sync throughout, but the four highest-value question topics in the data are still not answered.

article.cshtml · 9 accordion items · FAQPage JSON-LD kept in sync at every pass

The original five, with what each one is actually worth

1How does Maatdesk use AI in a legal environment? KD 1 — trivially winnable, and zero demand

2What types of legal tasks can AI support? KD 66 WITH NO VOLUME — WORST IN THE BLOCK

3How are AI tools used responsibly in law? KD 50, n/v · closest real kw is «closed ai systems for law firms» 40 vol

4How do AI tools integrate with cases and data? REPLACE WITH THE KD 11 QUERY THAT ALREADY RANKS AT POSITION 3

5Legal AI vs traditional software KD 24, 0 volume · already surfacing sitewide at position 24

The change log records the count (5 → 9) and that the accordion and FAQPage schema were kept in sync at every pass; it does not enumerate the four questions that were added, so they are not listed here. What is recorded about the new set: FAQ 9 is where «ai lawyer» (9,900 vol) and «ai attorney» (1,600 vol) appear — and they appear as a denial, because the product is not either of those things.

The four questions the data says to answer — all ranking, all at zero clicks STILL OPEN

Query	KD	Position	Impr	Status on the page
legal ai clause library migration	7	7.0	1	Lowest KD of any query in the dataset. Not on the page at all
migrate matters, contacts, documents to legal ai	11	3.0	4	Direct replacement for FAQ 4. Not on the page
custom legal ai implementation cost	10	7.5	14	Not on the page — and no pricing copy exists to support it
role based access legal ai	29	5.0	1	The page says «permissions» exactly once and never expands

These are not speculative targets. Google is already showing the site for all four, at positions 3–7.5, and returning zero clicks — because there is nothing on the page that answers them. Two of the four (migration and role-based access) also pair with the schema's unsupported claim of a «secure cloud-based SaaS platform», covered in the next section.

Links — outbound internal 1 → 9 to 7 unique pages · external 0 → 2 · inbound 4 → 7 pages, and inbound is now the constraint

The page had ONE internal link out and ZERO external citations — on a topic where the credible sources (the ABA and NIST) are exactly what a legal buyer and an AI answer engine both look for. It also had only 4 inbound internal links, and after this work inbound is the binding constraint on the page, not content.

article.cshtml · audited 2026-09-09 · outbound peaked at 18 during the work and was deliberately cut back to 9

6.1 Outbound internal — 1 → 9, to 7 unique pages +8

9 internal links across 7 destination pages — placed contextually, inside sentences that already named the product 2 external citations — ABA Formal Opinion 512 and the NIST AI Risk Management Framework, both in the client-data section

THE LINK BLOCK THAT WAS BUILT AND THEN DELETED IS THE FINDING HERE. A «Works with the rest of Maatdesk» block put 16 links in a single paragraph, 8 of them exact-match anchor text — a pattern that reads as a link farm to a reviewer and dilutes every anchor on the page. It was replaced with 4 contextual links in sentences that already named those products, taking outbound from 18 back down to 9.
Fewer links, placed in prose, was the correct outcome — and it also removed the exact-match anchors that would have competed with the pages owning those terms.

6.2 Inbound internal — 4 → 7 pages, and this is now the ceiling THE BINDING CONSTRAINT

Seven pages now link to /ai-tools, up from four. Content is no longer what limits this page. A 3,335-word page with 5 valid schema nodes and 7 inbound links is content-complete and authority-poor. Every further word added to this page has a smaller effect than one more relevant inbound link would. That is the recommendation the change log ends on, and it is unchanged here.

6.3 Cross-sell cannibalisation — checked, and the removal was the fix

The XSell terms — legal billing / calendaring / contact management / task management software — dropped to 0 on this page when the link block was removed. That is correct, not a regression. Each of those terms has its own landing page, and several already rank sitewide: «legal calendaring software» 374 impressions at position 32, «legal task management software» 434 at position 31, «legal contact management software» 227 at position 26. An AI page carrying exact-match anchors for all of them competes with its own siblings. The featureList was rewritten the same way — 6 AI capabilities instead of the platform's 13 modules.

JSON-LD & Entity Data — 3 → 5 nodes, rebuilt as one connected graph · 0 @id before, 5 after · two invalid Offer objects removed

The old markup was three disconnected nodes with no @id on any of them, no dateModified, and a generic photograph of a Miami office tower as the page image. It also carried a description asserting that the product drafts documents and summarises legal research. Google's 2026-09-09 crawl still reads that version — none of this is deployed yet.

Nodes 1–5

The graph — 3 nodes → 5, all cross-linked

Was NO @id ANYWHERE

SoftwareApplication (+3 Offers, featureList of 13 platform modules), VideoObject, FAQPage with 5 Q&A
No @id on any node, so nothing could reference anything else. No dateModified.
image = miami_tower_h.jpg — a generic building photo standing in for a software product

Now VERIFIED: 0 UNRESOLVED REFS

articlemetasSoftwareApplication + WebPage + BreadcrumbList

articleVideoObject + FAQPage (9 Q&A, up from 5)

wiringall five carry @id and are linked by about / isPartOf / mainEntityOfPage / breadcrumb

providerreferences the site #organization instead of duplicating it — same for publisher

dateModifiedwired to Model.GetLastModifiedDate, not hardcoded

Removed

priceSpecification — invalid on both paid Offers

What the test said

Google's Rich Results Test reported «Invalid object type» on both paid Offers. The UnitPriceSpecification was removed from each
The three Offers themselves are kept and correct: $0 six-month trial, $39/user/mo, $29/user/mo billed yearly

The part that is bigger than this page 46 OTHER FILES

46 OTHER FILES SITEWIDE still carry the same UnitPriceSpecification pattern and will be throwing the identical warning. Fixing them is a separate, larger job that nothing on this page depends on

Node 2

SoftwareApplication — featureList rebuilt

13 platform modules → 6 AI capabilities SCOPE CORRECTED

The featureList had been advertising Billing, Calendaring, Contact Management, Time Tracking, Task Management, Client Portal, Mobile App, E-Signature and Trust Accounting — the whole platform, on the AI page
Every one of those is an unqualified generic at KD 26–100, and every one has its own landing page that should rank for it
Now: the 6 AI capabilities this page actually describes

The claim still sitting in the description UNSUPPORTED BY VISIBLE COPY

The node asserts a «secure cloud-based SaaS platform». The new client-data section addresses this in part, but «role based access legal ai» (KD 29) ranks at position 5 sitewide and the page carries no role or permissions copy at all — the single «permissions» mention it used to have went out with the old FAQ answers

Node 4

VideoObject — and two earlier findings that were WRONG

Fixed duration WAS MISSING

durationPT43S — a Google-recommended property that was absent

inLanguageadded

contentUrladded

descriptionrewritten — it said the AI «transforms case management», a term /legal-case-management-software owns in its own title

skippedKey Moments (hasPart/Clip) — 43 seconds is too short for chapter markup

WITHDRAWN — two earlier drafts of this analysis were wrong CORRECTED 09-08

«The video is missing» — FALSE. ServiceEntities.xml sets vimeoslideid="1161936980" for id="ai-tools", matching the schema. The markup is valid
«The dates mismatch» — NOT A DEFECT. uploadDate describes the video, article:published_time describes the article. A 2024 page carrying a video uploaded in 2026 is normal
The 0 video rows in the GSC Search-appearance export are more plausibly explained by the page earning 0 clicks overall than by broken markup

Declined

aggregateRating — deliberately NOT added

Google's Rich Results Test on 2026-09-09 found 2 review snippets that exist nowhere in the source. The Elfsight widget in the footer injects them client-side.

A hardcoded rating would create two competing sources for the same claim — the worse failure mode, because the two would drift apart and neither would be authoritative.

The same call was made on the home page, for the same reason. If ratings are ever marked up, they should be pulled server-side, not typed in.

Open — sitewide

Duplicate @id values across three page pairs

/e-signature declares @id https://maatdesk.com/ai-tools#video — it has no vimeoslideid of its own, so it borrowed this page's identifier. Two different URLs now claim the same entity.

Two more duplicate @id pairs exist sitewide: legal-case-management-software#video and maatdesk.com/#video.

Not fixable from this page. Each borrowing page needs either its own video or its own @id namespace.

Deliberate Omissions & Measurement — 16,000+ monthly volume left on the table on purpose, and the three numbers that will judge this work

The largest keyword opportunities in this dataset were declined because the product does not do those things. «ai for legal research» (3,600), «lawyer ai» (3,600), «ai lawyer» (9,900) and the whole draft* family (1,580) are left at zero deliberately. The old meta descriptions claimed two of those capabilities, and removing the claims is what produced five of the negative deltas in the keyword table.

DO NOT «FIX» THESE — each one is a decision with a reason, recorded so the next editor does not undo it

What was left at zero, and why

«ai legal research» (720) · «ai for legal drafting» (140) · «legal drafting ai» (140) — LEFT AT ZERO. The product does not draft documents or perform legal research. The old meta descriptions claiming otherwise were removed for exactly this reason
«lawyer ai» (3,600) — LEFT AT ZERO. No sentence on this page places that word order naturally, and forcing it would read as stuffing
«ai lawyer» (9,900) and «ai attorney» (1,600) — appear ONLY in FAQ 9, as a denial. They are the two biggest AI terms in the gap set and the product is neither of those things
The XSell software terms dropped to 0 when the link block went. Correct: /legal-billing-software, /legal-calendaring-software, /legal-contact-management-software and /task-management-tool-for-attorneys should rank for them
aggregateRating NOT added — the Elfsight widget already injects review snippets client-side, and a second hardcoded source would compete with it

Technical changes that carry no keyword

sitemap.xml lastmod for /ai-tools: 2026-09-03 → 2026-09-09
scroll-margin-top: 110px on all 8 heading ids, for the sticky header
.ai-report* and .toc-card CSS added to custom.css AND main.min.css
ServicePageSkin.cshtml — text-lg on the brief-intro paragraph. Size only: .prose-text was rejected because it would also centre the paragraph and cap the measure at 68ch

How to judge this work — three positions, re-pulled in 4–6 weeks NOT WORD COUNT

Query	Position now	Impressions	What movement would mean
ai for legal software	45.5	S: 44	Largest sitewide AI query by impressions; Semrush shows 0 volume for it
ai for lawyers software	41.6	S: 39	Same cluster, same starting depth
matter management software ai	30.0	P: 7	The one on-target non-brand query this page already surfaces for

Re-pull the identical GSC export in 4–6 weeks and compare these three positions. Those numbers, not word count, will say whether the work landed.
AND THE CAVEAT THAT APPLIES TO EVERY FIGURE ABOVE: all impression and position data in this report was exported BEFORE the work, for 2026-06-07 → 2026-09-06. Nothing here is evidence that a ranking changed. The tables show what the edits did to the PAGE, not what Google did with it — and the changes are not deployed yet, so the clock has not started.

Source Index — 5 keyword data sets · 4 in-repo sources of truth · scan date 2026-09-08

Every figure in this report traces to one of the rows below

A. Keyword data

SR-GAPgap.keywords_2026-09-08T14_49_37.866Z.csv — 753 keywords. Semrush Keyword Gap: /ai-tools vs darrow.ai/resources/ai-tools-for-lawyers, legora.com and pro.bloomberglaw.com. The page ranks for 0 of the 753

SR-BULK1artificial-intelligence_mobile-app_bulk_us_2026-09-08_11-31-20.csv — Semrush Bulk Keyword Overview, US, 45 keywords

SR-BULK2legal-case-management-software_bulk_us_2026-09-08_11-42-06.csv — Semrush Bulk Keyword Overview, US, 36 keywords

GSC-PAGESearch Console, Page = https://maatdesk.com/ai-tools, Web, 2026-06-07 → 2026-09-06

GSC-KWSearch Console, Query = «AI Tools for Lawyers», same window — 0 impressions across all 92 days

GSC-SITESearch Console sitewide, same window: 1,000 queries / 15,509 impressions / 241 clicks

B. On-page sources of truth — the four files this page is actually made of

Views/ServicePage/Content/ai-tools/En/articlemetas.cshtml — metas + SoftwareApplication, WebPage and BreadcrumbList schema
Views/ServicePage/Content/ai-tools/En/article.cshtml — body copy + VideoObject and FAQPage schema
Views/ServicePage/ServicePageSkin.cshtml — renders the H1 and the brief intro
App_Data/ServiceEntities.xml, id="ai-tools" — the name attribute IS the H1, briefintro is the intro paragraph, and vimeoslideid is the video. Three of this page's most important elements are not in either .cshtml file

C. External, cited in the new client-data section TWO SOURCES

ABAFormal Opinion 512 — the profession's own guidance on generative AI

NISTAI Risk Management Framework

Nothing was asserted about where document text travels or what is retained, because that is documented nowhere. The section states what is sourced and stops.

Status — Done vs Open

NOT DEPLOYED. Everything described in this report is uncommitted. Google's 2026-09-09 crawl still reads the old structured data, including the description claiming the product drafts documents and summarises legal research. Until this ships, the measurement clock has not started.

PAGE STATE 2026-09-09, end of session. Body copy 3,253 words (255 at the start) · total indexable content 3,335 · full corpus incl. metas and schema 3,870 · 8 H2 sections (was 3) · 9 FAQ items (was 5) · internal links out 9 to 7 pages (was 1) · external citations 2 (was 0) · inbound internal links 7 pages (was 4) · schema 5 nodes, all @id'd, 0 unresolved references, 0 missing required properties · keyword occurrences 390 across 136 tracked rows (was 128).

Two edit sessions — 2026-09-08 and 2026-09-09 · every change mirrored as a Razor comment at the point of the edit

✓ Done

✓ Title rebuilt on «legal workflow automation» (480 / KD 22 / $52.22) after the bulk pull killed the earlier «AI Tools for Law Firms» recommendation (170 / KD 61)
✓ All three descriptions rewritten — and three false capability claims removed: document drafting, smart document generation, legal-research summarisation
✓ og:image and twitter:image were EMPTY; twitter:card was «summary». Both images set, card switched to summary_large_image, og:url made absolute
✓ Body copy 255 → 3,253 words across three passes; intro paragraph 0 → 78 words in briefintro
✓ 11 screenshots holding ~490 unindexable words converted to HTML report panels — then audited against the source images, and five transcription errors corrected
✓ H2 sections 3 → 8; four originals reworded to name Maatdesk without losing a keyword; an 8-entry «On This Page» index added
✓ Feature H3s kept as branding, with the real keyword moved into the subheads — matter management (KD 15), contract review (KD 29), legal document review (KD 21)
✓ FAQ 5 → 9, accordion and FAQPage schema kept in sync throughout
✓ Schema rebuilt into a connected graph: 3 → 5 nodes, 0 → 5 @id, dateModified wired to the model, featureList changed from 13 platform modules to the 6 AI capabilities
✓ VideoObject: duration PT43S, inLanguage and contentUrl added; description rewritten off a term another page owns
✓ Two invalid Offer priceSpecification objects removed after the Rich Results Test flagged them
✓ Internal links 1 → 9 across 7 pages, external citations 0 → 2 (ABA 512, NIST AI RMF); an 18-link version was built and cut back to 9 on purpose
✓ A 351-word section cut to 191 with zero keyword loss, because the sample panels above it already showed 14 of its 23 labels
✓ US spelling pass · sitemap lastmod → 2026-09-09 · scroll-margin-top for all 8 heading ids · report CSS added to both stylesheets

⚠ Open

⚠ NOT DEPLOYED — everything above is uncommitted, and Google is still crawling the old schema and the old claimsthis is the only item that blocks every other item
⚠ NO VISIBLE PRICING COPY. Four pricing queries rank at KD 3–31, positions 2.4–7.5, all at zero clicks; the three Offers exist only in schema«ai enabled practice management pricing» is KD 3 with Transactional intent at position 3
⚠ The four ranking FAQ topics are still unanswered — clause-library migration (KD 7), matter/contact/document migration (KD 11), implementation cost (KD 10), role-based access (KD 29)
⚠ /e-signature declares @id maatdesk.com/ai-tools#video — it has no video of its own. Two further duplicate @id pairs exist sitewide
⚠ The trial claim contradicts itself: «6 Month Free Trial» ×97 against «90-DAY RISK-FREE» ×12. The six-month version is the one in structured data, as an Offer at price 0a visitor and a crawler get different answers to the same question
⚠ 46 other files sitewide carry the same invalid UnitPriceSpecification pattern that was removed from this page's Offers
⚠ aria-expanded missing on all 9 FAQ accordion buttons. Accessibility only — the answers are in the DOM and fully indexed. Needs the shared accordion JS changed too
⚠ The privacy policy does not name OpenAI or Google Gemini as sub-processors. The new client-data section links to it, and a compliance reviewer following that link will not find AI processing described
⚠ Inbound links (7 pages) are now the binding constraint on this page, not content. Another 1,000 words would do less than one more relevant internal link
⚠ «maatdesk login» — 267 of this page's 302 de-anonymised impressions. /ai-tools should not be the result Google serves for a login querya routing and internal-linking problem, not a content one

TEST AFTER DEPLOY, in this order. Rich Results Test on the live URL — it confirms the 5 nodes parse and shows whether the Elfsight review snippets are still being injected. Search Console → URL Inspection → Request Indexing, because the currently indexed version still carries the removed claims. Then re-pull the same GSC export in 4–6 weeks and compare the three positions in section 8.
Standing instruction: every change made to this page is logged in ai-tools-improvements.txt as a matter of course, without being asked, and mirrored as a Razor comment at the exact point of the edit — with the help-centre or Semrush source it came from. Each pass appends its own section below the previous one: WHAT changed, the SOURCE relied on, the KEYWORDS reinforced with their volume and KD, and the updated open-items list carried forward. To refresh the table after the next round, re-parse both sides on the same day — «WAS» must keep meaning the git HEAD before 2026-09-08 no matter how many times the counter runs.

AI Tools for Lawyers — SEO Improvements