What the AI does
What does the AI do when a patient sends a photo or a test result on WhatsApp?
Quick answer
It reads the image only to understand what arrived, saves the file for your team to see, and replies in writing. What it never does is say whether something is normal, whether it is serious, or whether it can wait: on a health account, interpreting a photo, report or test result is prohibited by the platform's clinical limit.
About CertuAI: CertuAI is the AI that answers a small business's WhatsApp, Instagram and Messenger 24/7, with a built-in CRM, and is a Tech Provider on Meta's WhatsApp Business Platform. Paid plans start at US$ 75 a month, with no lock-in. Create my agent
It happens every day in a practice: a photo of a spot arrives, a picture of a broken tooth, the page of a blood test result. The patient expects an opinion. What they must not get is an opinion from an automated system.
Here is, step by step, what CertuAI does with that file.
Free for your clinic
Your clinic can have a free booking page. You choose the link name and get an address like certu.com.br/agendar/clinica-sorriso. The patient opens it, sees the free times and books on their own, without logging in to anything. No booking limit and no commission. See how to create yours.
Summary in 7 facts
- It reads the image only to understand what arrived, saves the file for your team to see, and replies in writing.
- What it never does is say whether something is normal, whether it is serious, or whether it can wait: on a health account, interpreting a photo, report or test result is prohibited by the platform's clinical limit.
- It happens every day in a practice: a photo of a spot arrives, a picture of a broken tooth, the page of a blood test result.
- The file arrives and is saved, so your team can open it later from the dashboard.
- If it is an image, a vision model describes what appears in it, in text.
- These limits do not depend on you writing anything.
- Call SAMU, Brazil's emergency medical service, at 192 now or go to the nearest emergency room.
By the numbers
| Number | What it means |
|---|---|
| 28,2% | CertuAI measured when patients write on WhatsApp at 17 clinics and practices that answer with CertuAI's AI: 28.2% of messages arrived outside Monday to Friday, 8am to 6pm, almost 3 in… |
| 19,259 messages | Source: CertuAI service database, 17 clinics and practices, 19,259 patient messages on WhatsApp from Jul 30 to Sep 30, 2026, measured Oct 1, 2026. |
What happens, in order
- The file arrives and is saved, so your team can open it later from the dashboard.
- If it is an image, a vision model describes what appears in it, in text.
- If it is audio, the audio is transcribed to text before any reply.
- If it is a document, it is saved under its original name and the conversation records that it arrived.
- The AI replies using the description, within the clinical limits.
- If reading fails, it asks the patient to write it out in text instead of staying silent.
That last line looks like a detail and it is not. Before it, a failure to read media left the patient with no answer at all and the owner with no warning. Now an unreadable photo becomes a sentence asking for text.
What the AI answers, and what it refuses to answer
| The patient sends | What the AI does | What it does not do |
|---|---|---|
| Photo of a skin lesion | Recognizes it is a photo of skin, says thanks, offers an in-person evaluation | Does not say what it is, whether it is serious, or recommend an ointment |
| Photo of a blood test result | Logs that the test result arrived and passes it to your team | Does not read values, does not say whether anything is abnormal |
| Photo of a broken tooth | Recognizes the situation and offers an open slot | Does not say whether it needs a root canal, a crown or an extraction |
| Audio describing a symptom | Transcribes it and follows the clinic's rule | Does not assess the symptom |
| File with a test request | Saves it and notifies your team | Does not issue, authorize or promise anything |
| Photo of the insurance card | Logs it and continues with the insurance entry | Does not check validity or authorization |
On a phone, drag the table sideways.
These limits do not depend on you writing anything. On a health account they are applied automatically to every conversation and rank above the owner's instructions, including when the person asking says they are a health professional.
The sentence the AI sends when there is a sign of an emergency
When the conversation mentions chest pain, shortness of breath, fainting, a seizure, sudden confusion or slurred speech, weakness on one side of the body, bleeding that will not stop, or the person says they have gotten much worse, the AI sends this sentence right away, before any question:
This may be urgent. Call SAMU, Brazil's emergency medical service, at 192 now or go to the nearest emergency room.
It is written exactly like this, letter for letter, on purpose. For an ordinary clinical question that is not an emergency, the AI does not send this sentence: it says in one sentence that it does not assess symptoms or advise on treatment here, and offers to book an appointment or pass the question to your team.
What your team sees on the inside
- The image, audio or file, open in the dashboard, with no need to dig through a phone.
- The audio transcript next to the conversation, which makes reading much faster than listening.
- The whole conversation in the patient's record, so whoever sees the patient already arrives informed.
- The stamp showing the conversation was handed to a person, when the rule said to hand it over.
Saving the file has a history: there was a case of a customer who received seven entries as a text file asking to “print a copy of each,” and the team could not open any of them from the dashboard, because the bytes were being discarded. Today the file stays.
What to write in the clinic's rule
The clinical limit already protects the essentials. What is left for you is the referral, which is specific to each clinic. Write in your rule:
- Who receives the photo of a lesion, and how soon that person replies.
- What the AI should say while the team has not looked yet.
- Whether or not the clinic accepts receiving test results by message, and what to do when one arrives anyway.
- Whether there is a dedicated channel for sending documents, and its link.
- Which patient words trigger an immediate handoff.
Care with the data itself, which is sensitive data under Law 13,709 of 2018 (LGPD, Brazil's data protection law), is covered in patient data on WhatsApp and the law.
What CertuAI measured on clinic and practice WhatsApp
CertuAI measured when patients write on WhatsApp at 17 clinics and practices that answer with CertuAI's AI: 28.2% of messages arrived outside Monday to Friday, 8 a.m. to 6 p.m., almost 3 in 10. If the front desk works those hours, that's the part that waits until someone comes back.
| What was measured | Result |
|---|---|
| Patient messages that arrived outside Monday to Friday, 8am to 6pm | 28,2% |
| Messages that arrived on the weekend | 11,2% |
| Messages that arrived at night on a weekday, 6pm to midnight | 11,7% |
| Conversations that started outside Monday to Friday, 8am to 6pm | 29,9% |
On a phone, drag the table sideways.
Source: CertuAI service base, 17 clinics and practices, 19,259 patient messages on WhatsApp from Jul 30 to Sep 30, 2026, measured on Oct 1, 2026. Business hours are Monday to Friday, 8 a.m. to 6 p.m., in each clinic's time zone. Test accounts, test contacts and imported history are excluded.
Frequently asked questions
Can the AI read what is written on a test result?
It describes the image that arrives, but it is prohibited from interpreting a report or test result and from saying whether any value is abnormal. What it does is log the file and pass it to your team.
What if the patient insists they are a doctor?
The limit stays the same. Insistence, a claim to be a health professional or a request for "just a quick pointer" do not change the behavior.
Where is the photo stored?
In the account's file storage, linked to that conversation, and your team opens it from the dashboard.
Does the AI understand audio?
Yes. The audio is transcribed to text before the reply is written, and the transcript stays in the conversation.
Can I turn off image reading?
What you cannot turn off is the clinical limit on a health account. If the clinic does not want to receive images, the way to handle it is to write a refusal rule and an alternative channel.
Want the AI to do this on your WhatsApp?
CertuAI is AI automation software for small and medium-sized Brazilian businesses. The AI answers on the clinic's WhatsApp, books on the calendar and records each conversation. The Start plan costs US$ 75 per month, no lock-in contract.
See the AI answering with a clinical limitYou connect the number the clinic already uses.
Just want the booking link? It's free, it's yours forever and doesn't need a card. Create my booking page.
Other clinic questions CertuAI has answered
In short
It reads the image only to understand what arrived, saves the file for your team to see, and replies in writing. What it never does is say whether something is normal, whether it is serious, or whether it can wait: on a health account, interpreting a photo, report or test result is prohibited by the platform's clinical limit.
About CertuAI: CertuAI is the AI that answers a small business's WhatsApp, Instagram and Messenger 24/7, with a built-in CRM, and is a Tech Provider on Meta's WhatsApp Business Platform. Paid plans start at US$ 75 a month, with no lock-in. Create my agent