The best live chat software for a therapy practice

Where therapy practices actually lose clients

Start with the leak, not the tool. A typical practice is one to three clinicians plus, when lucky, one admin. The clinicians are in session most of the day. The admin, when there is one, is scheduling, billing, and answering the phone. The prospective client is not in the building. The prospective client is on a phone at night, reading bios and trying to decide if this practice is safe to contact.

Two leak windows, not one

The first window is after hours. Reaching out to a therapist takes courage, and the moment of courage often arrives in the evening, when the visitor finally has privacy to look. A practice with only a phone number and a contact form has nothing to answer with in that moment. The second window is between sessions. Clinicians are with clients, the phone goes to voicemail, and a caller who was already nervous does not always call back. A chat widget covers both windows, because it answers when staff cannot, in either window.

The intake leak is a compliance leak too

Here is where therapy differs from every other vertical on this site. A visitor who writes "my name is Sarah and I need help with anxiety after my divorce" has just put protected health information into a chat transcript. If the chat vendor has not signed a business associate agreement, that transcript is a compliance problem, not a lead. A practice that buys the cheapest chat widget and lets it collect presenting concerns has bought a leak of a different kind. The rest of this page scores tools on both halves: how well they capture the inquiry, and whether they can legally hold what the inquiry contains.

The same ten questions, every day

Prospects ask the same questions over and over: do you take new clients, what are your fees, do you take my insurance, in person or by video, what are your hours, where are you located, how long is the wait, and which therapist fits my situation. The first seven are perfect AI questions, because the answers are stable and public. The last one, which therapist fits, is a matching question with a clinical edge, and it needs a careful script. The questions that are never AI questions are just as predictable: anything about a crisis, anything clinical, and anything that asks chat to confirm coverage. A practice that lets a bot guess at those three will harm people it meant to help. The rest of this page scores tools on both halves: how well they answer the repeatable seven, and how cleanly they hand the other three to a person.

What a therapy practice actually needs from chat

Eight requirements, drawn from how a therapy practice actually works. Every tool below is scored against these eight, and nothing else.

1. HIPAA and a signed BAA, or no PHI in chat

This is the hard line that decides the pick. A chat tool that can receive protected health information needs a signed business associate agreement, and the practice must confirm the plan it buys is covered by it. Among the eight tools on this page, only Intercom states HIPAA compliance on its own page, and only on its Expert plan. For every other tool, no HIPAA statement was verified from the pages checked, which means a practice buying one of them must either keep all personal detail out of chat or verify the vendor's BAA position itself before collecting anything personal. A vendor that does not state HIPAA compliance on its own page is not assumed to offer it. That is the rule we apply on this page, and it is why the cheapest tool does not win.

2. Crisis routing hard line

Chat must never become a crisis line. The greeting must carry a crisis notice, and any message that suggests self-harm, harm to others, or an emergency must surface emergency resources at once and go to a person, with no bot conversation in between. In the United States that means the 988 Suicide and Crisis Lifeline, by call or text, and 911 in an emergency. The tool needs keyword routing or a greeting that makes those resources visible before a visitor types anything clinical. A tool that cannot put a crisis notice in front of a visitor before the conversation starts fails this requirement outright.

3. Intake that collects facts, never clinical advice

Intake chat collects contact and logistics details: name, phone, email, preferred days, in person or by video, and whether the inquiry is for an adult, a teen, or a couple. It never asks for history, symptoms, or a diagnosis, and it never answers a clinical question. When a visitor volunteers clinical detail, the tool must acknowledge it, stop collecting, and hand the conversation to a person or to the secure intake form. The script matters more than the software here, but the software must let the practice write that script and must not improvise past it.

4. Therapist matching without steering care

Matching is the question every visitor asks and the one most likely to go wrong. Chat may share public facts: each therapist's specialties, populations served, session formats, languages, and availability. Chat must never state that a specific therapist is clinically right for a person, or that a modality will treat their condition. The safe move is a shortlist of two or three therapists whose public profiles fit the visitor's stated preferences, plus a person to confirm the fit. A tool that lets the practice load therapist bios as its knowledge base and answer only from those bios passes. A tool that lets its AI reason freely about treatment does not.

5. Insurance verification boundary

Insurance is the question that most often turns into a wrong answer. Chat may list the plans the practice is in network with, state the self-pay fee, and explain that benefits are checked after intake. Chat must never confirm that a specific plan covers therapy, quote a copay or deductible, or confirm eligibility. Those answers need a benefits check against the visitor's actual plan, and a bot that guesses at them creates a billing dispute and a broken first session. The boundary is simple: chat states the process, a person states the coverage.

6. Confidentiality in a shared inbox

A therapy practice inbox is not a shop inbox. An admin, a biller, and three clinicians may all need access, but not to the same messages, and not with the same visibility. The tool needs roles and assignment rules so a billing question does not sit open in front of every clinician, and a clinical handoff does not sit open in front of billing. Shared inboxes without roles are a confidentiality risk in a practice, not a convenience. This requirement also covers retention: chat transcripts are practice records, and the practice needs to know how long they are kept and who can read them.

7. After-hours capture for admin questions

The tool must answer fees, hours, location, formats, and new-client availability when nobody is at a desk: evenings, weekends, and every hour clinicians are in session. That means an AI answer product with a real allowance, not a chatbot demo. Allowances differ wildly, and they are the second most important number on this page after the BAA question. A practice should size the AI allowance at its real after-hours volume, not at the headline plan name.

8. Cost that fits a one to three clinician practice

Per-seat pricing multiplies fast for a small team, flat workspace pricing does not, and usage pricing depends on what counts. A practice should price the tool at its real team size and its real after-hours volume, not at the headline plan name. AI budgets are also counted in different units, and those units are not identical: Intercom charges per Fin outcome, LiveChat counts resolutions, Tidio counts only human replies as billable, Crisp budgets AI credits, and Freshchat counts sessions. The math section below prices each in its own unit. For therapy, the honest cost comparison starts at the HIPAA tier, not at the cheapest tier, because the cheapest tier is not the tier a practice that collects intake can legally use.

Comparison: all eight tools against practice requirements

Prices are annual billing unless labeled monthly, re-checked November 7, 2026, except where the method section notes an earlier check. HIPAA positions are stated only where the vendor's own page states them. "Not verified" means no HIPAA statement was found on the pages checked, not that the vendor cannot sign a BAA.

How to read the AI allowances. Intercom counts one Fin outcome per conversation. LiveChat counts a resolution as one solved question, and multiple questions in one chat still count as one. Tidio counts a billable conversation only when a human agent replies, so Lyro AI replies do not consume the quota, but Lyro conversations count against the separate Lyro quota. Freshchat counts a session as all AI interactions within 24 hours of the first message. Crisp counts automated conversations against an AI credit budget. Zendesk counts automated resolutions per agent. These are different units, so compare them as budgets, not as identical counts.

The pick: Intercom on Expert

Intercom wins on the requirement that decides a therapy purchase. Intake chat collects a name plus a reason for seeking therapy, and that is health information. Among the eight tools on this page, only Intercom states HIPAA compliance on its own page, and only on its Expert plan at $132 per seat a month billed annually. That is the whole pick in one sentence: it is the only tool here that clears the hard line on verified evidence. The pick is not about features or price. A cheaper tool that cannot hold intake is not a bargain for a practice. It is a compliance problem with a chat widget attached.

Expert at $132 a seat a month billed annually is expensive for a small practice, and we do not hide that. One seat is $132 a month billed annually, two seats are $264 a month billed annually, and Fin AI costs $0.99 per outcome on top. What the practice buys for that price is a plan tier the vendor itself describes as HIPAA compliant, plus Fin as the after-hours layer for admin questions, plus multibrand Messenger and Help Center on Expert for a group practice with more than one brand or location. The practice must still get a signed BAA in writing before launch, must confirm the plan terms cover its own use, and must configure Fin so crisis and clinical messages go to a person at once. A HIPAA statement on a pricing page is not a signed agreement, and no page on this site claims otherwise.

Two honest limits. First, the monthly Expert price of $139 is carried from an earlier read, not re-read on November 7, so a practice paying monthly should confirm it on the pricing page before budgeting. Second, Fin is a second meter, so a practice with heavy after-hours volume pays the seat price plus the outcome charges, and the math section prices that combination. Neither limit changes the pick, because no other tool on this page clears the hard line at any price.

Runner-up: Tidio, for admin-only chat

Tidio is the pick for a practice that can hold a strict line: chat answers admin questions only, and collects no personal detail. Fees, hours, location, formats, new-client availability, and a booking link are all fine in chat. The moment a visitor starts describing why they need therapy, chat stops collecting, shows the crisis notice when relevant, and hands the visitor to a phone call or the secure intake form. If a practice can hold that line, Tidio wins on price and channels. Its multichannel inbox covers website chat, Instagram, Messenger, and WhatsApp at no extra platform cost, with WhatsApp's only extra being the per-message rate above 1,000 free messages per number a month. Starter at $24.17 a month billed annually covers a workspace, not a seat, with up to 10 seats on self-serve plans, so a two-clinician practice pays the workspace price, not twice a seat price. Add Lyro from $32.50 a month billed annually for 50 Lyro conversations and the combined cost is about $56.67 a month billed annually. The free plan covers 50 conversations and a one-off 50 Lyro conversations, enough to test whether after-hours questions are real for a specific practice before paying. The 7-day trial needs no card.

The condition is the whole runner-up. Tidio's BAA position was not verified from the pages checked, so a practice that lets intake drift into presenting concerns in Tidio chat has no verified compliance basis for those transcripts. If the practice cannot hold the admin-only line, Tidio is not the right runner-up, and Intercom Expert is the only pick on this page. A practice should decide which of those two it is before it buys, not after the first real intake arrives in chat.

When a practice should pick LiveChat instead

LiveChat is a mature, dependable human-plus-AI tool, and its Growth plan is genuinely good: 200 AI resolutions a month, unlimited conversation history, multiple website support on all plans, and a 14-day trial with no card. For a typical single-location practice it loses on two counts. First, no HIPAA statement was verified from the pages checked, so a practice needing a BAA cannot assume one is available on these plans. Second, price. Instagram installs free from the marketplace, but WhatsApp costs $59 a month through a marketplace app, so a one-advisor practice on Growth pays $79 a month billed annually plus $59 for WhatsApp, $138 a month in total, and a two-person team pays $217 a month, against $132 a month billed annually for one Intercom Expert seat that carries a HIPAA statement. For a larger practice, a multi-location group, or a franchise with a dedicated admin team, that premium buys the most mature human chat tool of the ones here, with AI covering nights and staff handling days, plus agent groups for routing across locations. At that size the $138 or $217 is a smaller share of revenue and the workflow matters more than the lowest price. That is the boundary: small practice that needs intake cover, Intercom Expert; small practice that can stay admin-only, Tidio; larger operation, LiveChat.

Visit LiveChat

The LiveChat button above is an affiliate link. Every other tool link on this page is a plain link.

The other five, briefly

Crisp: flat pricing, good inbox separation, admin-only

Crisp is the pick for a practice whose chat volume is high and unpredictable. Conversations are unlimited on every plan, and pricing is flat per workspace, so a busy month cannot spike the bill. The free plan covers 2 seats with unlimited conversations, Mini at $45 covers 4 seats, and Essentials at $95 covers 10 seats and adds the omnichannel inbox with WhatsApp Business and Instagram DMs, plus 2 sub-inboxes for separating admin from clinical handoffs. The tradeoff is the AI budget: $5 of AI credits on Mini covers about 90 automated conversations, $25 on Essentials about 450, and $75 on Plus about 1,350. No HIPAA statement was verified, so Crisp is an admin-only option unless the practice verifies a BAA itself. Annual prices are not publicly shown, so budget monthly.

Freshchat: best for text-first practices, still admin-only on verified evidence

Some practices run on text. Clients ask about schedule changes by SMS, and prospects message on WhatsApp first. Freshchat's Growth plan at $19 a month billed annually per agent adds social messengers including Instagram plus SMS and WhatsApp in the same inbox, which none of the other low-priced plans bundle as plainly. Freddy AI Agent costs $49 per 100 sessions after a one-time 500 free sessions on paid plans, which is a workable AI budget for a practice testing automation. The catch is per-agent pricing: a three-clinician practice pays three times the plan price before any AI sessions, so compare it at team size, not at the headline rate. No HIPAA statement was verified, so the same admin-only condition applies.

Zoho SalesIQ: cheapest paid path, weakest AI story on the pages checked

Zoho's free plan covers 3 operators with 100 sessions, and Basic at $7 a month billed annually per operator adds a chatbot with 25,000 bot sessions, plus messaging channels including WhatsApp and Instagram listed on the pricing page. Professional at $12.75 a month billed annually adds translation and auto-translated bot content, which matters for a practice serving clients in more than one language. For a practice that mainly needs a staffed widget and a free start, it is the cheapest credible option. Its AI story on the pages checked is a chatbot builder and an Answer Bot on Enterprise, not a headline AI agent product, and no HIPAA statement was verified, so a practice buying mainly for after-hours automation or intake cover should weigh Intercom or Tidio first. Dollar prices are carried from October 29-30, not re-read November 7.

Zendesk: only if you already live there

Zendesk's $19 Support Team plan includes no live chat, so a practice buying for chat starts at Suite Team, $55 a month billed annually per agent, with monthly prices at $25, $69, and $149 in the comparison table. A practice already running Zendesk for support has a reason to stay. A practice buying chat alone does not, and no HIPAA statement was verified from the pages checked.

Olark: unlimited AI at an enterprise price

Olark's Standard AI Website Agent at $400 a month billed annually includes unlimited conversations with no token costs and no overages, plus one human seat, with extra seats at $29 a month and 2 agents confirmed at $429 a month in the FAQ. That flat unlimited model is genuinely attractive for a high-volume practice, and the human-only plan at $29 an agent a month is a cheap staffed fallback. At $4,800 a year, the AI plan is priced for a group or a very busy practice, not a single small practice, and no HIPAA statement was verified from the pages checked.

The math: what after-hours volume actually costs

Take a practice fielding about 100 after-hours admin questions a month. This is an example, not a statistic. Watch what happens to each tool's price at that volume.

Intercom Expert: One seat at $132 a month billed annually plus 100 Fin outcomes at $0.99 each is $99, for $231 a month billed annually. Two seats plus the same 100 outcomes is $264 plus $99, or $363 a month billed annually. The seat is the bill here, and the outcomes are the second meter. A practice should size Fin by its real after-hours volume, because every outcome past the seat price is $0.99.

Tidio: Lyro answers do not count as billable conversations, so the 100 AI answers cost nothing against the Starter quota. Starter at $24.17 a month billed annually covers the human-reply conversations, and the Lyro 50-pack at $32.50 a month billed annually covers the first 50 AI answers, about $56.67 a month billed annually combined. A practice at 100 AI answers a month needs a larger Lyro quota than the 50-pack, so budget above $56.67. The larger quota price is not stated on this page because it was not re-read from Tidio's own page in this run. Monthly, Starter at $29 plus Lyro at $39 is $68 a month.

LiveChat: Essential includes 10 AI resolutions. Covering 100 needs 2 extra 50-packs at $49.50 each, $99, for a total of $118 a month billed annually for one seat. Growth includes 200 resolutions for $79 a month billed annually, and is cheaper than Essential plus packs at this volume. The lesson is general: on LiveChat, buy the plan whose allowance covers the volume, do not bolt packs onto the small plan. Add WhatsApp at $59 a month and Growth for one advisor is $138 a month in total.

Crisp: Essentials at $95 a month includes about 450 automated conversations, which covers 100 with room to grow. Mini at $45 includes about 90, which does not cover 100, and Mini lacks the omnichannel inbox, so the realistic Crisp entry at this volume is Essentials, not Mini.

Freshchat: Growth at $19 a month billed annually per agent plus Freddy sessions: the first 100 sessions sit inside the one-time 500 free sessions on paid plans, then cost $49 per 100 sessions. A practice past its free sessions pays the seat price plus the session packs. Two agents on Growth are $38 a month billed annually before any Freddy packs.

Zoho: Basic at $7 a month billed annually per operator includes a chatbot with 25,000 bot sessions, which covers 100 questions many times over. Two operators on Professional are $25.50 a month billed annually. The price is real, but the product is a chatbot builder, not an AI agent, so compare it on answer quality, not on the session count.

Zendesk: Suite Team at $55 a month billed annually per agent is $110 a month for two agents, and Suite Pro at $115 is $230 a month for two agents. The automated resolution allowance is 5 per agent on Suite Team and 10 on Suite Pro, so a practice at 100 AI answers a month should expect overage charges on top, at $1.50 committed or $2.00 pay-as-you-go.

Price anchors from the suite side. A chat tool at $57 to $132 a month is cheap next to a practice management system, because it answers chat, not clinical work. No dollar figure for SimplePractice, TherapyNotes, or TheraNest is stated on this page, because none was verified from the vendor's own page in this run. Do not let a chat page sell a practice management system, or a system page sell chat as a replacement for staff. Budget for the leak actually measured: after-hours questions first, then the suite gap, if any.

Chat vs the clinical record: where one ends and the other begins

Many practices already pay for a system that includes some messaging. Buying a standalone tool on top of it can mean paying twice for the same inbox. Check the system first.

If you run SimplePractice, TherapyNotes, or TheraNest

SimplePractice, TherapyNotes, TheraNest, and similar practice management systems run scheduling, intake forms, clinical notes, treatment plans, insurance claims, telehealth, and the client portal. That is where the clinical record lives, and chat never holds it. If the system's messaging already covers the client portal, the standalone tool's job shrinks to the gaps: website visitors who have not yet become clients, Instagram DMs, or after-hours admin answers the system does not cover. No dollar figure for these systems is stated on this page, because none was verified from the vendor's own page in this run. Get pricing from the vendor before comparing system cost to chat cost.

If you run Calendly or a plain website

A practice on Calendly, Squarespace Scheduling, or a plain website has no system chat at all, and a standalone tool is the whole solution. In that case the widget's job is to capture the inquiry, answer the routine questions, and hand the visitor to the secure intake form with the fee, the insurance process, and the wait time already explained. That handoff is the product. Judge the tool on whether the handoff ever loses the visitor, not on how many features its pricing page lists.

Do not confuse chat with a crisis line or a clinical note

Chat tools answer written admin messages. Crisis lines answer people in crisis, and clinical notes hold the treatment record. A practice that needs crisis cover needs a crisis protocol and a person, not a chatbot. A practice that needs a clinical record needs its practice management system, not a chat transcript. The prices differ accordingly, and the measurement differs too: count which channel the missed inquiries come from before buying any of the three.

Setup checklist: the first week with chat on a practice site

This is the outcome work. A chat widget with an empty knowledge base is a prettier contact form. Load it with the practice's real answers before launch, and write the hard lines before any automation goes live.

Sort the BAA first, before any intake flows

Decide which of the two models the practice runs: intake chat under a signed BAA, or admin-only chat with no personal detail. If it is the first, get the agreement signed and confirm the plan terms cover the practice's use. If it is the second, write the admin-only rule into the tool and into staff training, and test it. Do not launch intake flows and sort the agreement later.

Write the crisis rule first, before any automation

The greeting carries the crisis notice, with the 988 Suicide and Crisis Lifeline by call or text and 911 for emergencies, visible before a visitor types anything. Any message suggesting self-harm, harm to others, or an emergency surfaces those resources at once and goes to a person, with no menu, no delay, and no AI attempt at counseling or risk assessment. Test this rule with example messages before launch, not after a real person in crisis hits it.

Write the intake script next to it

Chat asks for contact and logistics details only: name, phone, email, preferred days, in person or by video. It never asks for history, symptoms, or a diagnosis, and it never answers a clinical question. When a visitor volunteers clinical detail, the script acknowledges it, stops collecting, and hands the conversation to a person or to the secure intake form. Load the knowledge base with the practice's own pages: fees, insurance process, hours, location, formats, therapist bios, and the wait time.

Write the matching and insurance rules next

Matching shares public facts only: specialties, populations served, formats, languages, and availability, as a shortlist of two or three therapists plus a person to confirm fit. Insurance states the process only: which plans the practice is in network with, the self-pay fee, and that benefits are checked after intake. Chat never confirms coverage, never quotes a copay, and never states that a specific therapist is clinically right for a person.

Set the inbox rules and the after-hours path explicitly

Decide who sees what: admin sees scheduling and billing, clinicians see their own handoffs, and nobody browses the whole inbox by default. After hours, the widget answers routine questions, captures the visitor's contact details and preferred days, and points to the secure intake form with the next steps. It should not promise a callback time the practice cannot keep. A captured inquiry with an intake link beats a voicemail that nobody returns.

Measure three numbers for a month

After-hours questions answered without staff, inquiries that reached the secure intake form, and intake-to-first-session conversion for chat-started inquiries against phone inquiries. If the first number is high and the second is flat, the handoff is broken, not the widget. If the third is low, the intake experience needs work, not the chat tool.

Frequently asked questions

How can you check this page for yourself?