How can legal software companies win law firms and in-house teams through AI search?

Which law firms and legal teams buy software, and what is each one worth?

Three very different buyers: small firms, large firms and in-house legal departments, each with its own budget and process.

Small and mid-sized firms buy practice management, billing, document and intake tools, usually without an IT department. The ABA’s 2024 practice management report (opens in a new tab) found 65% of firms budget for technology, rising from 41% of solo respondents to 90% of firms with 100 or more attorneys, and an average annual technology spend of $13,991. Each customer is small, but there are many, and subscriptions renew.

Large firms buy legal research, document review, e-discovery and legal AI platforms through innovation teams, partners and security reviews. Their deals are larger and slower. Harvey’s claim that 80% of Am Law 100 law firms use it shows how concentrated and competitive the top of this market has become.

In-house legal teams buy contract management, matter management, e-billing and legal AI, often alongside procurement and IT. They are the heaviest AI users in Wolters Kluwer’s survey: 76% used generative AI weekly.

Investors are betting on the category. Harvey’s September 2026 round of $550 million at a $15.5 billion valuation is a self-reported figure, but it shows how much revenue the market expects legal AI to earn.

Where do AI assistants sit in how lawyers find software?

Already inside lawyers’ daily work, though no study isolates AI use for choosing legal software.

Lawyers know these tools first-hand. In the ABA’s 2024 AI survey (opens in a new tab) of 512 attorneys, 30.2% said their offices were using AI-based tools. Among AI research tools adopted or seriously considered, ChatGPT led at 52.1%.

How lawyers learned about new technology in that survey is telling:

That survey predates the spread of AI answers in search. Our inference is that the same sources, legal publications, legal news and peer firms, now also reach lawyers through AI answers that cite them.

Software buyers in general start more research with AI. G2 (opens in a new tab), a review platform, asked 1,076 software buyers in March 2026 where their research begins; 51% said they start with an AI chatbot more often than with Google. G2 sells visibility to software vendors, and its sample was not drawn from law firms or legal departments.

On Google, AI answers appear on most software searches. Across the 1,248 US searches in our AI Overview study, the B2B software and technology group, where legal tools belong, drew an AI Overview (the AI summary above Google’s results) on 96.0% of searches.

Which questions do legal buyers ask AI assistants?

Questions about fit, accuracy, confidentiality and ethics, as much as features. We wrote the example prompts below ourselves to show the pattern; none were collected from real lawyers.

To answer a prompt like the trust-accounting comparison, assistants can search the web. Google says AI Overviews and AI Mode may use a “query fan-out” technique (opens in a new tab) for such questions, and OpenAI says ChatGPT search rewrites a question (opens in a new tab) into targeted queries. Our hidden-searches study found ChatGPT running a search aimed at a named publication, ranking or award in 43.8% of its answers. For legal software, we infer, those would include legal technology press, bar association resources and independent evaluations.

How does an AI answer become a legal software subscription or contract?

Through trials for small firms and through pilots and security reviews for large firms and legal departments.

Small firms: answer, trial, subscription. A managing partner asks which practice management tool fits their practice area and size, gets a few names, and starts a trial. With no procurement team, the AI answer carries much of the shortlist, we infer.

Large firms: answer, pilot, firm-wide rollout. An innovation lead or partner researches legal AI tools, then runs a pilot that must pass security and ethics review. The answer decides who gets the pilot; the pilot decides the contract.

In-house teams: answer, evaluation, enterprise contract. A general counsel or legal operations lead frames options for contract or matter management, then runs a formal selection with procurement and IT.

Vendor diligence by lawyers is thinner than you might expect. In the ABA’s 2024 cloud computing report (opens in a new tab), only 23% of respondents said that, as a precaution when using cloud tools, they had evaluated the vendor company’s history, and 19% sought peer advice. If buyers check little themselves, what an AI answer says about you matters more, we infer.

The revenue will rarely appear as a tracked click. For ways to tie AI answers to signed firms and legal departments, see what lost clicks mean for pipeline.

Why would an AI assistant name one legal tech vendor over another?

No platform says how it picks; studies point to independent sources, and lawyers add accuracy, confidentiality and ethics.

Documented by the platforms. Google and OpenAI both describe their AI answers searching the web and citing what they find. Neither explains how it chooses which legal software vendors to put in front of a lawyer.

Observed in studies. Chen and colleagues (opens in a new tab) looked at US software questions, the nearest available proxy for legal tech, and found AI search drew 72.7% of its sources from earned media (independent reviews and publications), against 45.4% for Google, whose results leaned more on vendor sites.

What legal buyers weigh. Three trust factors are specific to this market:

• Accuracy. Stanford RegLab and HAI researchers tested leading legal research AI tools on over 200 legal queries and found that even they hallucinate in 1 out of 6 (or more) benchmarking queries (opens in a new tab): Lexis+ AI and Ask Practical Law AI gave incorrect information more than 17% of the time and Westlaw’s AI-Assisted Research more than 34%, still far less often than general chatbots. The study dates from 2024, and the vendors may have improved their tools since. In the ABA survey, accuracy was the top concern about AI tools, named by 74.7% of respondents.
• Confidentiality. In the ABA cloud report, confidentiality and security were the top concerns, cited by approximately 55% of respondents.
• Ethics. Formal Opinion 512 ties generative AI use to the model rules on competence, confidentiality, communication and fees. Stanford notes that by May 2024 more than 25 federal judges had issued standing orders on AI use in their courtrooms.

Our inference. Lawyers trust what other lawyers, courts and independent testers say. Assistants appear to lean on independent sources too. A reasonable expectation is that legal vendors with published security documentation, honest accuracy testing and coverage in legal publications give both readers more to work with. Nobody has yet tested this with practice management, research or contract tools.

What does it cost a legal software vendor to be missing or misdescribed?

Lost trials and pilots, and in this market, reputational risk from wrong claims. Nobody has yet priced that loss for legal tech.

• Small firms decide fast. Without procurement teams, a firm that gets three names from an assistant may trial only those, we infer.
• The top of the market is concentrated. With one vendor reporting use at 80% of Am Law 100 firms, a challenger missing from AI answers starts further behind.
• Wrong security or accuracy claims are costly. If an assistant says your product trains on client data when it does not, or misstates your certifications, you may be ruled out before a demo. Our guide on how to fix wrong brand information in AI answers covers corrections like these.
• One answer is not a verdict. Ask the same practice management question twice and the list may change: in our consistency study, only 25.2% of the brands ChatGPT named for a question appeared in all five runs.

What does GEO involve for a legal tech vendor?

It puts your accuracy, security and ethics evidence where AI assistants can find and quote it. No vendor can be promised a recommendation.

Generative engine optimization (GEO) means working to be mentioned, and described correctly, when AI assistants answer questions. For a legal software company it covers:

1. Clear identity by buyer. Separate pages for solo and small firms, large firms and in-house teams, stating practice areas, firm sizes and systems you integrate with.
2. Public trust documentation. Security certifications, data handling, whether client data trains any model, and data residency, on plain web pages, not only in sales decks. Confidentiality was the top cloud concern for lawyers.
3. Honest accuracy evidence. Publish how you test accuracy, take part in independent evaluations, and say plainly what the tool should not be used for. Overclaiming invites the scrutiny Stanford applied to “hallucination-free” marketing.
4. Ethics alignment. Explain how your product supports competence, confidentiality, supervision and billing under Formal Opinion 512 and state bar guidance.
5. Legal press, CLE and peers. Coverage in legal publications, CLE sessions and customer stories from named firms are the sources lawyers already learn from. Our piece on how brands build authority for AI search explains why those third-party sources carry weight.
6. Repeated checks for each buyer. Ask the small-firm, large-firm and in-house questions several times each in ChatGPT, Google AI Overviews and AI Mode, Gemini, Perplexity, Copilot and Claude, and track the answers.

For the wider software picture, see how B2B SaaS companies earn revenue from AI search, and for the limits of tactics, legitimate GEO versus manipulation. Law firms are not the only buyers who check a vendor’s handling of sensitive data before they buy: school districts run privacy checks too, as how EdTech companies win schools through AI shows.

What don’t we know yet about AI search in legal software buying?

Nobody has yet shown that appearing in AI answers wins more law firm subscriptions or legal department contracts.

• No study of legal buyers choosing software through AI. The ABA and Wolters Kluwer surveys measure AI use in legal work, not in vendor selection.
• Dated and interested sources. The ABA figures are from 2024, before AI answers spread in search. Harvey’s figures are its own; Wolters Kluwer and G2 sell to this market.
• Accuracy tests age quickly. The Stanford study tested 2024 versions of legal research tools.
• Revenue effects are the thinnest research. Across the 45 studies of AI search optimization reviewed by Martinez (opens in a new tab), traffic and conversions were the least supported area, the very outcome a vendor counting trials and contracts cares about.

How should a legal tech vendor test what AI tells lawyers about it?

Ask the assistants what lawyers ask before a trial or pilot, then check how they describe your safeguards.

Draft questions for solo and small firms, large firms and in-house teams covering practice fit, alternatives, accuracy, confidentiality and ethics. Put each one to ChatGPT, Google, Gemini, Perplexity, Copilot and Claude more than once, since answers vary between runs. Note which vendors appear, which legal publications or bar resources are cited, and whether your security posture, model training policy and integrations come out right. Where an answer is wrong or thin, the cause is usually proof you have not yet published.

To have us run that test with you, ask us for a review of your visibility to law firm and legal department buyers. We will show where AI answers place you for small firms, large firms and legal departments, and which missing public proof is most likely costing you trials, pilots and contracts. Publishing that trust, accuracy and ethics evidence, then checking how assistants repeat it, is the work our generative engine optimization service page lays out for legal tech vendors.

Frequently asked questions

Do lawyers use ChatGPT to choose legal software?

No study measures that directly. But ChatGPT was the top AI research tool firms had adopted or considered in the ABA’s 2024 survey, at 52.1%.

Does legal AI accuracy affect whether assistants recommend a tool?

Unknown, but it affects buyers. Stanford found even leading legal research tools hallucinated in at least 1 out of 6 benchmark queries in 2024.

Should legal software vendors publish their security details?

Yes, in plain web pages. Confidentiality and security were the top cloud concerns for about 55% of attorneys in the ABA survey.

Do ethics rules matter for legal software marketing?

Yes. ABA Formal Opinion 512 links generative AI use to competence, confidentiality, communication and fees, so buyers will ask how your tool supports them.

Is AI visibility more important for small-firm software?

Probably. Small firms rarely run formal selections, and only 19% of attorneys in the ABA cloud survey sought peer advice as a precaution with cloud tools.

Sources

• Wolters Kluwer (2024), 2024 Future Ready Lawyer Survey Report (opens in a new tab)
• American Bar Association (2025), 2024 Artificial Intelligence TechReport (opens in a new tab)
• American Bar Association (2025), 2024 Practice Management TechReport (opens in a new tab)
• American Bar Association (2025), 2024 Cloud Computing TechReport (opens in a new tab)
• American Bar Association (2024-07-29), ABA issues first ethics guidance on a lawyer’s use of AI tools (opens in a new tab)
• Harvey (2026-09-09), Harvey Raises $550M at a $15.5B Valuation to Help Legal Teams Own Their Intelligence (opens in a new tab)
• Stanford HAI (2024-05-23), AI on Trial: Legal Models Hallucinate in 1 out of 6 (or More) Benchmarking Queries (opens in a new tab)
• Damien Charlotin (2026), AI Hallucination Cases database (opens in a new tab)
• G2, Tim Sanders (2026-04-15), In the Answer Economy, Don’t Win the Click — Win the Answer (opens in a new tab)
• Google Search Central (2025), AI features and your website (opens in a new tab)
• OpenAI Help Center (2025), ChatGPT search (opens in a new tab)
• Chen, Wang, Chen and Koudas (2025), Generative Engine Optimization: How to Dominate AI Search (opens in a new tab), arXiv:2509.08919.
• Martinez (2026), Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023-2026) (opens in a new tab), arXiv:2607.14035.
• Underneath (2026), When does Google show an AI Overview? 1,248 US searches
• Underneath (2026), The hidden searches AI assistants run before they answer
• Underneath (2026), Ask an AI the same question 5 times: do the brands change?