Criminal defense firms get named after an arrest by winning the second and third questions, not the first. An answer engine reads “I was arrested last night” as a request for procedure and answers it with procedure, then closes with a generic line about contacting a criminal defense attorney. The firm that supplied the procedure answer is the one still in the conversation when the reader asks who to call.
Everything published about which law firms answer engines name is personal injury data. No equivalent criminal defense dataset exists, and the agency pages currently competing for this topic carry percentages with no study attached. This post separates what has been measured from what is inferred from adjacent data, and labels which is which.
- Answer engines treat “I was arrested last night” as a request for procedure, not for a lawyer, and answer it with booking, bail and arraignment steps followed by a generic instruction to contact a criminal defense attorney.
- No published study measures which criminal defense firms AI engines name. Citorian’s five-engine test of June 2026 is personal injury data, and its finding that situational phrasing produced a named firm 18% of the time against 94% for direct vetting asks is a read-across, not a criminal defense measurement.
- Whitespark’s Q2 2025 local study found AI Overviews on 15% of local-intent queries but 97% of hybrid queries, the kind that add a charge, a county and a prior conviction, which is how a frightened family member actually types.
- The winnable surface for a defense firm is the jurisdiction-specific procedure page the engine quotes, because that is what is still present when the same person asks who to call.
Why does an answer engine describe the process instead of naming a defense lawyer?
Because the question as typed asks for the process. An arrest question carries no hiring intent, so the engine returns booking, bail, the arraignment window and the right to counsel, then hands off without a name. Citorian’s five-engine test of June 2026, covering 359 answers, found situational phrasing produced a named firm 18% of the time against 94% for direct vetting asks.
That gradient is personal injury data. Citorian ran ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews across five US metros between June 13 and June 15, 2026, and tested no criminal defense prompts. Firm AEO carries the numbers across because the mechanism is the phrasing rather than the practice area, and treats the criminal defense figures as illustrative until somebody publishes a test. Any agency page stating a criminal defense naming rate as fact is stating a number nobody has measured.
The generic handoff sentence is the opening. When an engine writes that a person should contact a criminal defense attorney in their county, the engine has already retrieved and often cited the page it took the procedure from. A citation is not a recommendation, but a cited firm is inside the conversation at the moment the reader has no name and needs one.
Which arrest questions can a criminal defense firm actually win?
The questions that carry a charge and a jurisdiction. A bare situational question is answered from general legal knowledge, which no firm owns. A question naming a county, an offense level and a prior conviction forces the engine to retrieve local material, and local material is written by local firms. The table below maps the sequence one person moves through in a few hours, from the arrest to the hiring decision.
| Question as typed | What the engine returns | Naming likelihood | What a defense firm can own |
|---|---|---|---|
| “I was arrested last night, what happens now” | Booking, bail, the arraignment window, right to counsel | Lowest band, situational phrasing | The cited procedure page for that county |
| “What happens after a DUI arrest in Maricopa County with a prior” | Local procedure, penalty ranges, sometimes a source list | Low to middling, hybrid phrasing | The page that states the local rule as a number |
| “Do I need a lawyer for a first offense DUI in Phoenix” | An advice-shaped answer, occasionally a firm | Middling, question-led phrasing | An answer-shaped page written for the person deciding |
| “Best criminal defense lawyer in Phoenix for DUI” | A short list of named firms | Highest band, direct vetting ask | Directory profiles and the review record |
| “Is this firm any good for a felony charge” | A summary assembled from third-party sources | Near certain naming, the firm is the subject | Entity consistency and third-party corroboration |
Who is typing the question at two in the morning?
Usually not the arrested person. Somebody in custody does not have a phone. The search is run by a parent, a spouse or a roommate who has just taken a call from a county jail, and that person knows the name, the county and almost nothing else. Content written in the second person for the accused misses the reader who is actually holding the device.
Location is now part of that retrieval. OpenAI enabled opt-in device location sharing in ChatGPT on March 26, 2026 across iOS and the web, which draws local answers from Google Business Profile, Apple Business Connect, Bing Places and directory listings rather than from prose alone. A firm with an unclaimed or inconsistent listing is absent from that layer regardless of how good its content is.
Detail is what puts an AI Overview in the path. Whitespark’s Q2 2025 study of 540 queries across Houston, Phoenix and Denver found AI Overviews on 15% of local-intent queries, with the local pack appearing on 93%, while hybrid queries that combine a location with a qualifying circumstance showed AI Overviews 97% of the time and the local pack only 17%. A frightened relative types hybrid queries by default, because the circumstances are the only thing they know.
One correction to the material circulating in this lane. Several criminal defense marketing pages assert that 78% of legal searches trigger an AI Overview. That figure is SE Ranking’s 77.67% for the legal niche inside a 1,200-query YMYL study, and SE Ranking’s broader 100,013-keyword research put legal at 0.34%. The spread is driven by query mix, Firm AEO has reconciled the competing trigger-rate studies separately, and no published dataset timestamps arrest-related queries at all. The overnight pattern is a reasonable working assumption from intake logs, not a measured fact.
What has to exist on the site before an engine quotes a defense firm?
A page that answers one charge in one county, with the local rule stated as a number. Engines retrieve passages, not websites, and a passage only survives extraction if it holds its meaning alone. The Princeton generative engine optimization study by Aggarwal and colleagues, presented at KDD 2024, found statistics and citations raised visibility in generative engines by roughly 22% to 41%.
The numeric version of a sentence outperforms the descriptive one, and depth matters more here than in most practice areas because the charge list is long and the procedure is county-level. Ahrefs found in March 2026 that only about 38% of pages cited in AI Overviews rank in Google’s top ten, down from 76% in mid-2025, so a firm sitting on page two for a charge-plus-county query is not out of the running for the citation.
- One page per charge per county, not one page for criminal defense, since the retrievable unit is the charge and the jurisdiction together.
- The governing rule written as a figure, for example the hours between booking and first appearance in that county, with the source named in the sentence.
- A sentence that reads correctly when quoted with nothing around it, naming the firm, the county and the charge type rather than using pronouns.
- A section addressed to the family member calling on behalf of somebody in custody, which is the reader most of these pages ignore.
- Claimed profiles and a live review record, since SOCi’s 2026 Local Visibility Index found ChatGPT recommending 1.2% of locations and the recommended ones averaging 4.3 stars.
What a criminal defense firm cannot put in that content
Anything that reads as a promise about the outcome. ABA Model Rule 7.1 prohibits false or misleading communications about a lawyer’s services, and a page built to be quoted by a machine is exactly where a line such as charges dismissed can be lifted out of its qualifier and presented as a claim.
The safest construction states what the process allows rather than what the firm achieves. Any case result carries its qualifier inside the same sentence rather than in a footer disclaimer the engine will never carry.
Intake automation is the second exposure, and it is sharper in criminal defense than elsewhere. Florida Bar Advisory Opinion 24-1 permits AI chatbots for intake and advertising subject to four conditions: the user is told they are dealing with AI, the tool gives no legal advice, represented persons are screened out, and the design avoids creating an accidental attorney-client relationship. A person who was arrested hours ago may already have appointed counsel, which makes the screening condition operational rather than theoretical. ABA Formal Opinion 512 of July 29, 2024 applies the competence, confidentiality and supervision duties to any AI-drafted marketing copy as well.
One page worth publishing is the one explaining that a conversation with a chatbot about a pending charge carries no privilege, a question consumers are now asking directly and Firm AEO addresses on its own. Firm AEO is a marketing company rather than a law firm, so everything above is a constraint on how content is built. The firm’s ethics counsel decides what the rules permit in its jurisdiction.
Will ChatGPT recommend a criminal defense lawyer at all?
It will, when the question is phrased as a hiring question. Citorian’s five-engine test found direct vetting asks produced a named firm 94% of the time in personal injury, against 18% for situational phrasing. An arrest question is situational by nature, so the realistic goal is being cited in the procedure answer and named in the vetting answer that follows it.
Is there any published data on which criminal defense firms AI engines name?
None that Firm AEO has found. Every published naming study, Citorian’s included, covers personal injury. A defense firm that wants a figure for its own market has to generate it: build a list of arrest-stage and vetting-stage prompts for its county, run each three times in a logged-out session on each engine, and record how often any firm is named and which firms they are.
Does a 24/7 phone line make an engine more likely to name the firm?
No engine publishes availability as a factor. Opening hours are structured data on Google Business Profile, which Gemini and Google AI Overviews read, so accurate hours help the listing behave correctly. The stronger argument for the overnight line is that the person reading the answer at 3 a.m. acts within minutes, which is a conversion question rather than a visibility one.
Should a defense firm write one DUI page or a page for every charge?
A page for every charge that the firm genuinely handles, split by county where the procedure differs. The retrievable unit is the charge plus the jurisdiction, and a single broad page competes for every variant at once while holding the specifics for none of them.
Can a firm advertise that ChatGPT recommended it?
Treat that claim as unusable. ABA Model Rule 7.1 turns on whether a statement is verifiable, and an answer engine produces different output across sessions, accounts and locations, so no third party can reproduce the result. Describing the work as improving how a firm appears in AI answers is accurate; describing the appearance as an endorsement is not.
- 1.Citorian, Which personal injury lawyers does AI actually recommend? Five-engine study (June 15, 2026)
- 2.Whitespark, The prevalence of AI Overviews in local search (Q2 2025)
- 3.Search Engine Land, ChatGPT enables location sharing for more precise local responses
- 4.SE Ranking, AI Overviews and YMYL topics research
- 5.SE Ranking, Google AI Overviews research (100,013 keywords)
- 6.Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)
- 7.Ahrefs, Only 38% of AI Overview citations rank in the top 10
- 8.Search Engine Land, SOCi 2026 Local Visibility Index
- 9.Florida Bar Advisory Opinion 24-1, generative AI in intake and advertising
- 10.ABA Formal Opinion 512, Generative Artificial Intelligence Tools (July 29, 2024)
- 11.ABA Model Rule 7.1, Communications concerning a lawyer’s services