Privacy Policy

• We ask for as little as we can: an email address, the state you are in, and whatever you choose to tell us about your animal.
• We never publish a conversation, we never sell personal information, and we never use your conversation to train an AI model.
• A conversation is read by you and by the professional who takes your case. It is not a public page and it never becomes one.
• We keep a permanent record of the exact words you were shown before any charge. That record includes the IP address the request came from and your browser’s description of itself, because it is the evidence that a disclosure was actually shown to you. It is used for nothing else.
• There is one piece of another company’s code on the public pages: a Google advertising tag, which tells Google that a click on one of our ads led to a purchase and what the purchase was worth. It does not receive your conversation, your email address or anything you typed about your animal. There is no analytics product and no advertising network of our own.
• Our own help chat bubble, in the corner of the public pages, opens a support session for your browser as soon as a page loads, whether or not you ever use it. That session records which pages of this site you open, your browser and system, the site that sent you here and, if you arrived from an ad, the click id in the address. It is how our support staff see who is on the site. It never holds your conversation with a professional. The section on what we collect says exactly what is in it and how long we keep it.
• You can ask for a copy of everything we hold, ask us to correct it, or ask us to delete it, by writing to hello@talktoavet.com.

That summary is not the policy. The rest of this page is, and it is more specific.

Your email address

Asked for once, at the point where you acknowledge what this service is and is not, which happens before you are shown a price. It is how we send your sign-in link, how we tell you your case has an answer, and how we send the reminder before a charge. There is no password on this service, so your inbox is the key to your account. We do not ask for your name.

The state you are in

Two letters, asked at the start and confirmed again before payment. We need it because who may answer your question depends on where your animal is: we cannot take questions from Texas, North Carolina, Alabama or Minnesota at all, and Virginia, Colorado and Idaho need a professional licensed in that state. We store the rule that applied on the day you paid alongside your case, so that a later change to our own rules cannot rewrite what governed your conversation.

What you tell us about your animal

Name, species, breed, age, weight, what changed and when, medications and known conditions: whatever you type. The intake assistant asks one question at a time, and your answers are kept as part of the case, together with the written summary the professional reads before replying. If you start a conversation and never pay, we still hold what you typed.

The conversation itself

Every message, from you, from the intake assistant and from the professional, with the time it was sent. This is the case record. It is what you come back to during your follow-up week, what the professional works from, and what a records request produces. We also record, for our own quality review, whether a professional’s reply was drafted with the help of an automated tool before they edited it and sent it. That marker is never shown to you and never goes outside the company.

Photos and documents you attach

Up to five files on a message: JPEG, PNG, WebP, HEIC or PDF, up to 10 MB each. Your browser uploads them straight into private storage using a link that expires, so they never pass through our application and never sit at a public address. Opening one later needs a fresh link that lasts five minutes, and only you or the professional holding your case can ask for one.

The help chat bubble, before you use it

The small chat bubble on the public pages is customer support for the service, not a place to ask about your animal, and it is the one part of this site that records a visit before you have done anything. When a public page loads, your browser opens a support session with our server so that the people working the support desk can see who is on the site and what page they are looking at. That session holds: a random id for your browser (so that a support conversation you start is the same one when you come back); the path of the page you are on and of the pages you open after it, up to the last fifty, with the time (the path only, never the search or the address bar’s query, and never a page of your conversation); a description of your browser, its version, your operating system, whether the device is a phone, tablet or desktop, and your screen size; the site that sent you here, as a site name only, never the page you were reading on it; the advertising click id and campaign labels (gclid, fbclid, utm_*) from the address you first arrived on, if there were any; the address the request came from; a heartbeat while the tab is open, so the desk can tell whether you are still here; and your name and email address only if you are signed in, in which case the session is tied to your account. It is kept apart from your case: nothing you or a professional writes in a conversation ever travels into it, and nothing in it is used to decide what you are charged or what you are told about your animal. If a member of staff blocks an address for abuse, that address is kept on a block list so the bubble refuses it; the block list is never applied to the paid conversation.

To keep that session the same one across visits, the bubble stores four entries in your browser’s local storage, listed by name in the cookies section. It uses no cookie of its own. If you do write in it, your messages travel over the same realtime provider as the paid conversation and a picture you attach goes to the same storage provider, both named in the section on the companies that help us run this; the session record itself goes to no other company.

Your phone number, only if you buy a call

We ask for it at the moment you buy a phone call, and not before, and we keep it so that the call can be connected. The professional never sees it: they are dialled first, and what shows on your phone is a TalkToAVet number rather than theirs. Your number is never displayed in the professional’s console.

How you rated the conversation

A rating out of five and, if you write one, a comment. The rating decides whether a bonus is owed on that conversation, so the professional can see it afterwards on their own pay statement. It is never put in front of them while they are answering you. Your written comment comes to us and is not displayed to the professional anywhere in the product.

Payment events from our processor

Your card details never reach us. Payment is taken on the processor’s own page, and what comes back to us is the event: a charge, a refund, a failed payment or a dispute, with the amount, the time, the processor’s reference for it, and an identifier that tells us which account it belongs to. We do not hold your card number and we could not produce it if you asked us to. We also keep the state of your membership, when the next charge is due, and the record that a reminder was sent before it.

How many conversations you have had

On a membership we count your completed conversations over a billing period. The count is worked out from your own cases rather than being separate information about you, and it exists so that the fair use policy is applied to numbers instead of to impressions. It is also how we know whether the price of the service still works.

Consent records, and why they carry an IP address

When you acknowledge what this service is, and again when you agree to a charge, we write a record that points at the exact wording that was on your screen at that moment, and stamps the time, the IP address the request came from, and the user-agent string your browser sent.

Those two technical facts are kept for one purpose: they are the evidence that a particular disclosure was shown to a particular person at a particular time. If you, your bank or a regulator later asks what you agreed to, we can answer with the words themselves rather than with a description of them. We do not use your IP address to profile you, to target advertising, or to work out where you are beyond the state you told us. We never edit one of these records: a change of wording creates a new version and leaves the old one exactly as it was.

You can read your own copy of this history on your account page, in the same form we would hand to anyone else entitled to ask for it.

• We never publish a conversation. Not as a page, not as an example, not with the names taken out. Confidentiality here runs to you, the client, and the fact that no veterinarian-client-patient relationship is formed does not release your transcript. Where this site writes about a condition, a person writes it from scratch.
• We never sell personal information. Nobody pays us for information about you, and no part of your conversation is used to advertise anything to you or to anyone else. We do buy advertising ourselves, and the measurement that goes with it is described in the section on advertising measurement below, in full, including what it leaves in your browser.
• We never use your conversation to train an AI model. We also choose providers whose own published terms say they do not train on what is sent to them. What we have not done yet is sign our own written agreement with each of them holding them to it. Until we have, this is the honest version of that promise and we will not print a stronger one.
• We never give the professional a direct line to you. They do not get your email address or your phone number, and you do not get theirs.
• We never put your words in a text message. When we page a professional about a waiting case, the text says a case is waiting; they read the case in the console, behind a login.

• You. Through the link you land on when you come back from payment, and after that by signing in with an emailed link.
• The professional who takes your case. While a case is waiting to be picked up it can be read by the licensed professionals who are eligible to answer it, because that is how someone decides to take it. Once it is taken it is open to the one person holding it. A professional who did not take it cannot open it, and neither can a professional who handled a different case of yours: they see the titles and dates of your earlier cases, not the conversations inside them.
• Us, for support and audit. The consoles our professionals use give nobody a route into a conversation they are not holding, an administrator included. But the people who run the company can reach the database those conversations live in, because someone has to be able to answer “what happened on my case” and to check that the service does what this page says. That access sits with those people, and it is for support and audit rather than for reading.

Two things you authorise at the start are worth repeating here. We review conversations for quality and safety. And where the professional uses an automated tool to draft a reply, the conversation is sent to our model provider for that purpose; the professional edits the draft and is responsible for every word that actually reaches you.

We buy advertising on Google. To know which ads are worth buying we run Google’s advertising tag on the public pages of this site. It is the only script from another company that this site loads, and this section says exactly what it does, because the honest description is shorter than the excuses would be.

What it stores in your browser

If you arrive from one of our ads, the address you land on carries a click identifier, and Google’s tag copies it into a cookie called _gcl_aw on this site’s own domain — sometimes one or two others alongside it, all with names beginning _gcl_, depending on how the ad was served. Their whole purpose is to remember, for up to ninety days, that this browser came from an ad, so that a purchase made later can be matched back to the click. They are set by Google’s script rather than by us, which also means, unlike our own cookies, that code running in the page can read them.

The tag is loaded on every page rather than waiting for you to click something, because the click identifier is only in the address bar at the moment you land.

Our own server keeps a copy of that click identifier too, in a cookie of ours called ttav_attr, and writes it against the conversation you go on to pay for. The reason is that Google’s tag only ever sees the browser the ad was clicked in and only ever sees the first payment. You may click an ad on a phone and sign in from an emailed link on a laptop; you may be refunded a week later, or dispute a charge. So, from our server rather than your browser, we send Google a short list of (click identifier, time, amount) rows when a payment lands, when a membership renews, and when money goes back the other way, so that a refund counts against the ad that produced the sale. Nothing about you goes in those rows: no name, no email, no state, no animal, and none of what was written in the conversation. The cookie lives 90 days because that is the window in which Google will still accept a row for a click, and it is described again in the cookies section.

What we send, and when

Five events, and nothing else. Each one says that something happened on this browser, and Google matches it to the click. None of them carries your email address, your phone number, your state, your animal, or any part of what you typed.

• When the assistant accepts your first message and a conversation starts.
• When the intake finishes and you are shown the price.
• When you go through to the payment page.
• When a payment completes, with the amount charged, and with a one-way digest of your case reference so that a refresh of the page is not counted as a second sale. Like every Google tag, this event (and the page view before it) also carries the address of the page it fired on, and because a completed payment is reported from your conversation page, that address contains your case reference. The reference is a random string that opens nothing without the signed key in your own browser or your sign-in, and Google is sent none of what the page says.
• When the assistant decides that what you have described is an emergency and sends you to a clinic. We would rather say this plainly than leave it out: that event tells Google an emergency triage happened on this browser. It carries no amount and no case reference, we never bid towards it, and we read it for one purpose, which is to find the searches that only ever end that way and stop paying for them.

Your browser also keeps a short marker for each event it has already sent, so that reloading a page does not report the same thing twice. Those markers live in your browser’s session storage, they are gone when you close the tab, and they hold a case reference and the word “sent”, nothing more. The only other thing this site keeps in your browser’s storage belongs to the support chat bubble, and it is listed with the markers in the cookies section so that nobody reading only that section is misled. The bubble keeps its own copy of the advertising click id from your first landing, for the support desk’s context; that copy is never sent to Google and never used to report a sale. What Google is told comes from the tag on this page and from the cookie in the cookies section, and from nothing else.

What the tag is not

There is no analytics product on this site, no session recording, no heatmap, no social pixel, no advertising network of our own and no profile built from what you read. (The help chat bubble keeps a list of the pages you open, described in what we collect; that is a list of paths for our own support staff, not a recording of what you did on them, and it goes to no one else.) The tag does load on the page your conversation lives on, because that is where a completed payment is reported, but it does not read a word of what you or the professional writes there. From that page it reports that it was opened (the page’s address, which holds your case reference and nothing else about you), that a payment completed, and what it was worth; on that page the tag is also told not to use the visit for advertising personalisation, so your conversation page can never place you on an advertising audience list. The tag does not load at all on your account page, on the professionals’ console or on our administration pages.

None of this runs while the site is behind its pre-launch password, and none of it runs if the advertising account is not configured.

We use as few as we can. Every one of them except the last handles information because we asked it to, on our instructions, for the job described here, and not for its own purposes. The last one is Google, and it is different in exactly that way, so it is described as what it is rather than tucked in with the others.

• Anthropic, our model provider. Gets the intake conversation. This is the part people usually skip, so plainly: when the assistant asks you the next question, and again when it writes the summary the professional reads, the words you have typed so far are sent to a model run by another company and its answer comes back to us. The same happens if the professional asks for a draft reply. It is processing, not publication, and none of it is public, but your words do leave our servers for it to happen.
• Resend, our mail provider. Gets your email address and the message we are sending you: a sign-in link, or a reminder before a charge. It does not get your conversation.
• Twilio, our text and voice provider. Gets a professional’s mobile number when we page them about a waiting case, and, if you buy a phone call, both numbers so that the call can be connected. It never gets any part of your conversation.
• Ably, our realtime provider. Carries messages as they are sent, so the other side’s screen updates without a refresh. The text of a message does pass through it, on a channel scoped to a single case that only you and the professional holding it can listen to. With it switched off, the pages simply check for new messages every few seconds instead.
• Our object storage provider. Holds the photos and documents you attach. The storage is private: there is no public address for a file.
• Stripe, our payment processor. Gets your email address, and your card details, which you type on their page and not on ours. It sends back the payment events described above.
• Google, for advertising measurement. The odd one out on this list, and named here rather than left off it. Its tag runs on the public pages, it sets the _gcl_aw cookie described in the section on advertising measurement, and it receives the five events listed there: that a conversation started, that a price was shown, that a payment page was reached, that a payment completed and what it was worth, and that a case was triaged as an emergency. It gets no part of your conversation, no email address and no phone number. Unlike the companies above it, Google uses what it receives for its own purposes as well as for ours, which is exactly why it is called out separately here instead of being folded in with the rest.

The application and its database run on infrastructure we operate ourselves rather than on a shared platform.

Your conversation

Kept, so that you can come back to it and so the professional can answer a follow-up during your week. After that week you can ask us to delete it, which is what we told you at the start and what we will do. There is no clock on it today: nothing in the software deletes a conversation by itself.

Consent records and the wording behind them

Kept, and never edited. These are the evidence of what you were shown before we charged you, and the reason to hold them is the same reason you might want us to: if there is ever a question about a charge, the answer is the words that were on the screen. Adding to that record is the only thing we ever do to it.

The record of what professionals are paid

Kept, and also added to rather than edited: a correction is a new line. It records which professional was paid what for which case, it holds no part of a conversation, and it is the accounting record of money owed to a contractor.

The help chat session

The support session the chat bubble opens, and any support conversation in it, are kept until you ask us to delete them or a member of staff removes them; there is no clock on them today either. Ask for them with your account or on their own, quoting nothing more than the rough time and the pages you were on if you were never signed in, and a person removes them. The block list of addresses refused by the bubble is kept until a member of staff lifts the block.

Cases we turned away

If you write from a state we cannot serve, or if the assistant decides that what you are describing is an emergency, we keep the case. You are shown no price and you are never charged, and the record is our evidence that we sent you to a clinic or turned the question away rather than taking money for it. You can ask us to delete it, the same as any other.

Write to hello@talktoavet.com from the email address on your account and say what you want. Because there is no password on this service, the address on the account is how we know it is you; if you write from another address we will ask you to confirm from the one on the account. We reply within ten business days, and we do not charge for any of this.

• A copy of what we hold. Some of it you can read yourself, right now, on your account page: your email address, your state, your membership, the list of your cases, and the full history of what you agreed to and when. Ask us and we will send the rest, including the conversations themselves.
• Correction. If your email address, your state or something you told us about your animal is wrong, tell us and we will change it. Two things we cannot rewrite, and would not: a message that was already sent, and a consent record. Those are the record of what happened.
• Deletion. Ask, and we delete. What that means in practice is the next section.
• The records of your conversation. Every answer you get carries the professional’s name, credential, license state and number, how to ask them for records, and where to complain about them. A records request goes through the instructions on that card, in the professional’s own words, because the record is theirs to hold and to hand over. If they have not written their own instructions, the card tells you to email us, quoting the case reference, and we will get the request to them.

We apply all of this to anyone who asks, wherever they live, rather than only where a statute compels it.

There is no delete button in the product today. You write to us and a person does it, which is slower, and worth saying plainly rather than implying that a switch exists.

What goes: your email address, your phone number if we have one, the state on your account, your sign-in links, the content of your conversations, including any photos you attached, and the help chat bubble’s session and any support conversation in it.

Two things stay, and you should know that before you ask.

• The record of what you were shown and agreed to before a charge, with its time and its technical stamps. Deleting that would leave charges we could not account for, to you or to anyone else.
• The accounting line showing that a professional was paid for a conversation. It names the professional and the case rather than you, and it is money that has already changed hands.

There is a third thing we should be straight about. A professional can download their own copy of a conversation they held, as their own professional record, and they are entitled to keep it. We cannot delete their copy for them. A request about that copy goes to them, using the instructions on the answer card.

Five of them are ours, and each does a job the site cannot do without. Ours are http-only, meaning code running in the page cannot read them, and in production they are sent only over HTTPS. None of ours advertises to you or follows you to another site; one of them, the last of the five, remembers which of our ads you arrived from, and it is described plainly below. Once we are advertising there is a sixth, and it is not ours.

• Your sign-in. It holds a signed reference to your account and nothing else: no name, no email address, no history. Thirty days, then you sign in again with a fresh emailed link.
• Your key to one conversation. One per case. It is set when you come back from payment, before you have had any reason to sign in, so that the page you land on is yours rather than open to anyone who guesses a link. It is signed, it lasts eight days, and a key to one conversation cannot open another.
• A professional’s sign-in. The same as the first, for a veterinary professional signing in to the console. Never set for a pet owner.
• The pre-launch door. Set only while this site is behind its pre-launch password, and only on a browser that typed it. It holds a digest rather than the password, and it goes away when the password does.
• Which ad brought you. Called ttav_attr. Set only when you arrive from a link that carries an advertising click id or campaign label (gclid, gbraid, wbraid, utm_*), it stores that click id, the campaign labels, the site of the page that linked to you, the page you landed on and the time, for 90 days. It is http-only and it identifies a click, not you. We use it to tell Google, from our server, that a click led to a purchase, a renewal, a refund or a dispute, so that refunds are counted against the ad as well as sales. It contains none of your words about your animal and no contact details.
• Google’s advertising cookie, called _gcl_aw. Written by Google’s tag, on this site’s domain, if you arrived from one of our ads, and it holds the identifier of that ad click for up to ninety days. It exists for one purpose, which is matching a purchase back to the ad. It is not http-only, so page code can read it, and it is the one cookie here that does the thing the other four do not. Depending on how the ad was served, Google’s tag may set one or two others whose names also begin with _gcl_, for the same attribution purpose. What they are for and what we send alongside them are in the section on advertising measurement.

Two more things that are not cookies. While you are using the site, your browser’s session storage holds a short marker for each advertising event already reported, so a page reload does not report it twice; those markers are gone when you close the tab. And the support chat bubble keeps four entries in your browser’s local storage: ttav_support_anon_id, a random id so that your support conversation is the same one when you come back; ttav_support_source and ttav_support_referrer, the kind of place and the name of the site that first sent you here (the site only, never the page); and ttav_support_ads, the advertising click id and campaign labels from the address you first arrived on, if there were any. They hold no name, no email address and nothing about your animal, and they stay until you clear the site’s data. What the server keeps against that id is in the section on what we collect, under the help chat bubble.

This service is for adults. We do not ask for information about children, we have no reason to collect any, and we do not knowingly collect personal information from a child under 13. If you believe a child has given us something, write to hello@talktoavet.com and we will delete it.

Not a promise that nothing can go wrong. Just what we actually do.

• There are no passwords on this service. Signing in is a single-use link that expires in fifteen minutes, and the database stores only a one-way hash of it, so a copy of our database could not be used to sign in as you.
• Attachments sit in private storage and are only ever reachable through a link that expires in five minutes, minted for the person entitled to see that case.
• Every credential this site issues is signed and scoped: a key to one conversation is useless against another, and an owner’s key is useless as a professional’s.
• A phone call is bridged through our own number in both directions, so neither party’s number is disclosed to the other.

We will change this page as the service changes, and the date at the top is the date of the version you are reading. If a change affects how we handle your conversation, we will ask you again rather than read wording you agreed to months ago as covering something new. That is also how the record works: a changed word makes a new version, and the old one is kept exactly as it was, still pointing at what you agreed to on the day.

Write to hello@talktoavet.com about anything on this page. If your question is about a particular conversation, quote the case reference and write from the address on your account.

Complaints about a professional go to the state board that licensed them. The card on every answer names that board and links to its complaint page.