For twenty years, a person with a legal problem did roughly the same thing: they typed it into Google, clicked through several firm websites, picked one, and filled out a contact form. Search was the front door, and the entire discipline of law-firm marketing was built to win a ranking on that door.
That front door is shifting. A growing share of people now begin somewhere else entirely — by describing their problem to an AI assistant like ChatGPT, Claude, Gemini, or Perplexity and asking it what to do. This is still an emerging behavior, not the majority path, and any honest description should say so. But the direction is clear, and it is worth getting in front of rather than behind.
Two things are changing at once. First, people are asking assistants to answer questions that used to start a search. Second — and this is the part most firms have not planned for — those assistants are beginning to act on the person's behalf: researching firms, comparing them, and, increasingly, initiating contact. When the first move in a legal matter is made by software, both how a firm gets found and how a firm receives an inquiry have to change.
Answer-engine visibility is the new search ranking
Answer-engine optimization, or AEO, is the practice of making a firm's facts clear, structured, and quotable enough that AI assistants represent the firm accurately when a person asks. It is the successor to SEO, and it rewards different things.
Search rewarded pages: keywords, backlinks, freshness, the mechanics of ranking a URL. Answer engines reward entities. When someone asks an assistant "who handles construction-defect cases in this county, and are they taking new clients," the assistant is not ranking pages — it is assembling an answer from whatever facts about your firm it can find and trust. If those facts are ambiguous, buried in an image, or contradicted across the web, your firm is easy to leave out of the answer.
In practice, AEO means a few concrete things for a law firm. State the firm's facts plainly and machine-readably: practice areas, jurisdictions, the attorneys and their bar admissions, and whether the firm is currently accepting clients. Keep those facts consistent everywhere they appear, because an answer engine that finds three different versions of your practice areas trusts none of them. And write in clear, quotable statements — definitions, direct answers, plain claims — because the sentences an assistant can lift cleanly are the ones it repeats.
The uncomfortable implication: a firm can have a beautiful website and still be nearly invisible to an answer engine, because the site was built to impress a human reader, not to be parsed and quoted by a machine. Entity clarity is now part of discoverability.
Receiving an inquiry from an AI agent
Getting quoted accurately is only half of it. The larger shift is transactional: an assistant that can gather the facts of a matter and submit an inquiry on the person's behalf needs a firm that can receive that inquiry, check it, and hand it to a human safely. Most firms' intake — a contact form and a voicemail box — was never designed for a well-formed request arriving from software at two in the morning.
Agent-first intake is intake designed to accept a structured inquiry from an AI assistant, verify it with the actual human, and route it into the firm's normal, attorney-governed process — without ever letting the software practice law or create a relationship the firm didn't agree to. This is what Intakit's agent-intake capability is built to do, and it rests on a few deliberate principles.
Discovery the agent can read: Each firm publishes a machine-readable identity — verified firm facts, attorneys and admissions, practice areas, jurisdictions, and whether it is accepting clients — served from records the firm itself owns and edits, not AI-generated guesses. This is the discovery layer an assistant reads before it ever tries to make contact.
A structured intake channel, not an open door: When an assistant submits on someone's behalf, the submission passes a locked allowlist of fields, rate limits, and per-firm quotas before a single AI cycle is spent. The firm accepts a well-formed inquiry, not an open pipe.
Nothing is real until a human confirms it: Every agent-submitted intake is held pending. The actual person receives a secure link, reviews exactly what the assistant said on their behalf, corrects anything wrong, and confirms their own identity. Consent to be contacted — including by text message — is granted only by that human, never inferred from an agent's claim.
No advice, no relationship, by design: Disclaimers travel with every interaction, in both human- and machine-readable form: submitting an intake does not create an attorney-client relationship, and no surface in the agent channel gives legal advice. Where the assistant answers at all, it stays to process facts.
Screening behind an information barrier: Conflict screening stays entirely on the firm's side of the wall. An inquiring agent never sees the firm's client data, party names, or the reason a firm declines — status responses to the agent are deliberately coarse, and the firm's actual conflict review happens internally, by the firm's own staff, after a human has verified the inquiry.
Minimize and record: Unverified agent intakes are purged automatically after a short window, consistent with how the profession treats prospective-client information under the ABA Model Rule 1.18 standard. State-changing actions carry a tamper-evident, hash-linked audit entry with agent-source attribution, so the record of what happened exists if a regulator, court, or the firm ever asks.
The through-line is that the agent does the midnight fact-gathering and the firm's humans make every professional judgment in the morning. The agent channel feeds the same hardened intake pipeline every other inquiry uses; it is a new front door, not a new set of rules about who gets to decide.
What a firm should do now
The honest framing is that this is an early channel, not a solved market. Agents are not the majority of legal inquiries today, and no firm should rebuild its practice around a channel that is still forming. But early channels reward the firms that show up before they are crowded, and the work of showing up is not speculative — it is mostly hygiene the firm should want anyway.
Make the firm quotable: publish clean, consistent, machine-readable facts about who you are, what you handle, where, and whether you are taking clients. Then make sure the intake can receive a structured inquiry, verify it with a human, and route it into attorney-governed follow-up rather than dropping it into a voicemail box that empties overnight. A firm that does both is discoverable when a client's assistant goes looking, and ready to catch the inquiry when the assistant reaches out — which is the whole game as the front door moves.