A business law firm wins a 'firm A versus firm B' prompt by being the side of the comparison an engine can verify. ChatGPT, Perplexity and Google AI Mode answer comparison prompts mostly from live third-party sources: Chambers and Legal 500 profiles, deal announcements, attorney bios and press. The firm with specific, checkable facts in those places gets described accurately.
This is the first Firm AEO post on business and corporate practice, and the first on comparison prompts as their own category. Earlier posts covered how engines pick a firm to name from scratch. This one covers the narrower case where the prospective client already has two names and asks the engine to choose.
- Graphite's August 2026 test found 58% of logged-out ChatGPT answers to entity-comparison prompts used web search, against roughly 7% of all U.S. ChatGPT responses carrying citations in Similarweb data, so comparison prompts are far more likely to be answered from live sources.
- BuzzStream's June 2026 study of 30,000 citations found evaluative prompts such as 'Is X better than Y?' had the highest rate of all four engines citing the same source, nearly three times the rate for brand-awareness or exploratory prompts.
- The FTI Consulting and Relativity General Counsel Report 2025 found 44% of general counsel say their legal teams use generative AI, up from 28% in 2024, and 79% say they turn to outside counsel more frequently.
- ABA Model Rule 7.1 comment [3] says an unsubstantiated comparison of a firm's services or fees with other firms may be misleading, so the software-industry 'us versus them' comparison page does not transfer to law firms.
- In the 5WPR and Haute Lawyer April 2026 audit, Chambers and Legal 500 profiles outranked Cravath's own pages for Cravath's own M&A practice, which shows where engines look when they compare business firms.
Why do comparison prompts behave differently from 'best business lawyer near me'?
Comparison prompts trigger live search far more often than ordinary prompts, so the answer depends on what is published now rather than on what a model remembered. Graphite's August 20, 2026 research found 58% of logged-out ChatGPT responses to 100 entity-comparison prompts used search, while Similarweb data cited in the same paper puts the share of all U.S. ChatGPT responses with citations at about 7%.
Engines also agree with each other more on comparisons. BuzzStream's June 10, 2026 study of 30,000 citations across 595 prompts on ChatGPT, Gemini, Google AI Overviews and Google AI Mode found evaluative prompts produced the highest rate of all four platforms citing the same source, nearly three times the rate for brand-awareness or exploratory prompts. The absolute rate was still under 1%, so agreement is relative, not common. Either way, the firm with thinner evidence is the one described in generalities.
Who asks an AI engine to compare two business law firms?
Two kinds of buyer ask comparison prompts about business firms: the founder or owner choosing counsel for a financing, sale or contract dispute, and the in-house general counsel checking a shortlist. Both usually arrive with two or three names from referrals and use the engine to separate them, which is why comparison prompts sit late in the buying process.
The in-house side is growing. The FTI Consulting and Relativity General Counsel Report 2025, based on 207 surveyed general counsel and 34 interviews, found 44% say their legal teams use generative AI, up from 28% in 2024, and 79% say they turn to outside counsel more frequently. No published study measures how many general counsel use AI specifically to compare outside firms, so Firm AEO treats that share as unknown.
The closest buyer data comes from software, not law. G2's 2025 Buyer Behavior Report, a survey of 1,169 B2B decision-makers fielded in April 2025, found 29% start research in tools like ChatGPT more often than Google. That is a read-across from software purchasing and should not be quoted as a legal-market figure.
What evidence does an AI engine use to compare two business law firms?
An engine compares business firms on the dimensions it can find written down by someone other than the firm: directory rankings and the commentary behind them, named transactions, attorney experience, industry focus and team size. The 5WPR and Haute Lawyer Legal AI Visibility Index, April 2026, found Chambers and Legal 500 owned multiple positions on elite corporate queries.
The 5WPR audit's clearest example is a query about Cravath, Swaine and Moore's M&A expertise, where two Chambers profiles, a Legal 500 profile and a Chambers Global profile sat alongside only two Cravath practice pages. Chambers says its research runs on about 350,000 interviews and surveys and 62,000 submissions a year, with client and peer feedback behind each ranking. That outside commentary is the kind of sentence an engine can lift into a comparison.
| Dimension the buyer compares | Where the engine usually finds it | What the firm can publish |
|---|---|---|
| Standing in the practice | Chambers, Legal 500, Best Lawyers commentary | Accurate directory submissions; the rankings as awarded, with year and body named |
| Deal or matter experience | Press releases, deal announcements, legal press | Matter lists with client consent, stated by type, industry and approximate size band |
| Industry fit | Attorney bios, client industry pages, conference speaking | One page per industry served, naming the transaction types handled |
| Team depth | Firm attorney directory, LinkedIn, bar records | Current bios with admissions, years in practice and practice focus |
| Fee structure | Rarely published by anyone | Billing models offered (hourly, fixed fee for defined work, caps), without comparing to other firms |
| Responsiveness and service | Directory client quotes, reviews | Nothing self-asserted; this dimension is earned in third-party commentary |
Can a law firm publish its own 'us versus them' comparison page?
Generally not in the way software companies do. ABA Model Rule 7.1 comment [3] says an unsubstantiated comparison of a lawyer's or law firm's services or fees with those of other lawyers or firms may be misleading if it is specific enough that a reasonable person would think it could be substantiated. A page titled 'Firm A versus Firm B' is that kind of comparison.
That closes the main tactic the software industry uses for comparison prompts. It does not close the prompt. A firm can still publish the facts a buyer would compare, about itself only, in specific and verifiable form: the transaction types it handles, the industries it serves, the billing models it offers and the rankings it holds with the year and the ranking body named. The engine does the comparing. The firm's ethics counsel decides how any individual claim reads under the state's version of Rule 7.1.
Matter lists carry a second constraint. Client identity and deal details are confidential under Model Rule 1.6 unless already public or the client consents, so the usual sources are deals the client has announced and matters described by type rather than by name.
Which facts should a business law firm make easy to extract?
The facts that win comparisons are the ones that can be quoted alone without losing meaning. A sentence such as 'the firm's corporate group handles venture financings for software and life sciences companies from seed through Series C' gives an engine a usable comparison point. A sentence such as 'we deliver exceptional results for growing businesses' gives it nothing.
- 1. One page per transaction type (venture financing, M&A sell-side, M&A buy-side, commercial contracts, entity formation), each opening with the work the firm does and for whom.
- 2. Attorney bios that state admissions, years in practice and representative work by type, kept consistent with the firm's directory profiles.
- 3. Rankings stated exactly as awarded, with the ranking body, band and year, and no restatement as 'top' or 'best'.
- 4. Billing models described plainly, without any claim of being cheaper or faster than other firms.
- 5. Deal announcements published on the firm's site when the client has announced the transaction, so the engine finds the firm's name next to the deal.
How should a business law firm measure comparison prompts?
Measure comparison prompts as their own set, separate from discovery prompts, and record how the firm is described, not only whether it is named. In a comparison the firm is almost always named because the buyer typed its name. The useful outcomes are whether the description is accurate and specific and which sources the engine cited for each side.
Run each prompt logged out, three times, on ChatGPT, Google AI Mode and Perplexity, monthly. The prompt shapes below are illustrative; a firm builds its own from the competitors its intake team actually hears named. The cited-source column is the work order: a competitor described from Chambers commentary while the firm is described from its own homepage points to the directory submission, not the website.
| Prompt shape | What to record | What a weak result points to |
|---|---|---|
| '[Firm] vs [competitor] for a Series A financing' | Accuracy of each description; sources cited for each firm | Missing transaction-type page or thin deal record |
| 'Which is better for selling my company, [Firm] or [competitor]?' | Whether the engine names a deciding fact, and whose | No sell-side M&A page; no announced deals tied to the firm |
| '[Firm] or [competitor] for a SaaS company's commercial contracts' | Whether industry fit is stated for the firm | No industry page; bios silent on industry |
| 'Is [Firm] good for startups?' | Whether directory or press commentary is cited | Absent or outdated directory profile |
| 'Compare fees at [Firm] and [competitor]' | Whether the engine invents a fee comparison | Billing models unpublished, leaving the engine to guess |
Should a business law firm block AI engines from comparing it to competitors?
A firm cannot stop a buyer from typing two names into an engine, and blocking search crawlers only removes the firm's own pages from the evidence. The engine then compares using directories and press alone. Keeping the site crawlable and accurate is the safer position.
What if ChatGPT gets a fact about our firm wrong in a comparison?
Record the answer and the cited sources, then correct the source the engine relied on: an outdated directory profile, an old bio or a stale press page. Engines that search live usually reflect the correction after the page is recrawled. Answers given from model memory change only with a new model release.
Can a firm mention its Chambers or Legal 500 ranking in its own content?
Most firms do, and accuracy is the constraint: state the ranking body, the band or tier and the year exactly as awarded, and do not restate it as 'best' or 'top'. Several states add conditions on referencing third-party ratings, so the firm's ethics counsel should approve the wording.
Do comparison prompts matter for smaller business law firms without Chambers rankings?
Yes, because the buyer who names a smaller firm already has a referral and is checking it. Without directory commentary, the engine leans on the firm's own pages, bios and announced deals, which makes specific transaction-type and industry pages more important, not less.
- 1.Graphite (Gregory Druck), Do not force search in prompt tracking (August 20, 2026)
- 2.BuzzStream, Which prompt types get the most AI citations? (June 10, 2026)
- 3.FTI Consulting and Relativity, The General Counsel Report 2025 (March 26, 2025)
- 4.G2, 2025 Buyer Behavior Report (May 14, 2025)
- 5.5WPR and Haute Lawyer, Legal AI Visibility Index 2026 (April 2026)
- 6.Chambers and Partners, Research methodology
- 7.ABA Model Rule 7.1, Communications Concerning a Lawyer's Services
- 8.ABA Model Rule 1.6, Confidentiality of Information