Can a WhatsApp AI assistant understand photos, screenshots, or PDFs?

Yes. A whacl assistant can work with supported photos, screenshots, and PDFs sent in a WhatsApp conversation. It can use information it reads from the file alongside your catalog and policies, then reply in the same conversation.

For example, a customer can send a photo of a damaged item, a screenshot of a payment error, or a PDF invoice and ask a question about it. Instead of asking the customer to type out information already visible in the attachment, whacl can use the file as part of the conversation.

What whacl can understand

Customers can send media through WhatsApp, Instagram, or Messenger, and the conversation appears in the same whacl inbox. Depending on the file quality, the assistant can read visible text, identify relevant details, and connect them to the business information you have provided.

• A product photo showing a model, color, size, label, or visible damage.
• A screenshot of a checkout, login, delivery, or payment error.
• A photo of a meter reading, serial number, receipt, or handwritten note when the text is clear enough to read.
• A readable PDF quotation, invoice, instruction sheet, application form, or shipping document.
• A file that answers part of a question while your catalog or policies provide the rest of the answer.

The attachment is considered in the context of the conversation. If a customer asks about a return and then sends a photo of the item, whacl can use both the image and the return policy you supplied.

Product photos and damaged-item claims

Suppose a customer writes, “This arrived broken,” and sends a photo. whacl can identify visible damage and follow the next step in your configured policy, such as asking for the order number, requesting additional evidence, or handing the conversation to your team.

It should not make a warranty decision from a blurry image or invent a policy exception. If the image is unclear, it can ask for a clearer photo. If your policy requires human review, whacl can hand over the conversation with the customer's messages and attachment context intact.

The same applies to product identification. If a readable model or product label matches your catalog, whacl can provide the catalog information. If it cannot identify the product confidently, it can ask for a model number or pass the request to your team instead of guessing.

Screenshots in support conversations

Screenshots can contain the exact error a customer cannot easily describe. A customer might send a payment failure screen, a shipping-tracking page, a booking confirmation, or a message from your website. whacl can read visible text and use it to determine what the customer is asking about.

For example, if a screenshot shows that a discount code was rejected, whacl can compare the visible message with your promotion rules and explain a likely reason, such as an expiry date, an excluded product, or a minimum order value.

Reading a screenshot does not give the assistant access to the system shown in it. Your team still controls refunds, account changes, order cancellations, and manual approvals unless those actions have been connected separately. When an action requires internal access, whacl can collect the relevant details and hand the chat to a team member.

PDFs and business documents

Customers may send a quotation, invoice, delivery document, or instruction sheet because that is where the relevant information already exists. When the PDF is readable, whacl can use its text to answer questions about line items, reference numbers, amounts, sections, or next steps.

For example, a customer asking, “Does this quote include installation?” can receive an answer based on the quotation and your business information instead of being asked to type out every line.

Large, scanned, password-protected, or poorly formatted PDFs may limit extraction. Tables, stamps, handwriting, and very small text can also require review. In those cases, whacl can ask the customer to resend a clearer file or hand the conversation to your team with the original attachment available in the thread.

Your catalog and policies control the business answer

Media understanding is not a substitute for current business information. The attachment might show a product code, while your catalog supplies the price. A screenshot might show a return request, while your return policy determines what the customer can do. A PDF might list a delivery date, while your policy explains what happens if that date has passed.

If the required information is not in your catalog or policies, the safer response is to ask a targeted question or involve a person. Changing the model will not fix missing prices, duplicate product names, or an unclear refund approval process.

Review mode and automatic replies

You can use review mode while testing media-based conversations. The assistant prepares a response for a team member to check before it is sent. This is useful for warranty claims, documents containing personal information, unclear product photos, and workflows where a wrong answer has a cost.

You can also use auto-send for media questions your assistant handles reliably, such as a clear screenshot of a known checkout error or a readable product label that matches your catalog. Review mode and auto-send are settings for the same whacl assistant, so you can expand automatic replies as your team reviews real examples.

Choose the model for your media questions

whacl lets you switch between Claude, GPT, and Gemini from one setting. You can compare how they read product labels, handle screenshots, summarize PDFs, and follow your returns policy without creating a separate inbox for each provider.

The best choice depends on the files and policies your business actually receives. Test with representative customer files, and keep in mind that model selection does not replace clear catalog data or well-defined business rules.

When a human should take over

A blurry serial number, disputed invoice, legal notice, medical document, or refund request may be too sensitive or consequential for an automatic answer. A handoff is also appropriate when whacl cannot read the file confidently, cannot match a product to your catalog, or reaches a policy boundary.

When whacl hands over a conversation, your team can see the full thread, including the customer's messages and attachment context, in the same inbox. They can review what the assistant understood, identify what is missing, and continue without asking the customer to repeat the issue.

Track the cost per reply

Media questions can involve more processing than short text questions, especially when the assistant reads a long PDF or combines an attachment with a detailed catalog lookup. whacl tracks cost per reply so you can compare the cost of text replies, screenshot troubleshooting, product-photo questions, and PDF-based requests.

That information can help you decide which media workflows are suitable for automatic replies and which are better sent to a human.

What customers should send

Clear files and specific questions give whacl more to work with. Customers can include an order number, product name, or short explanation with the attachment. For screenshots, the full error message is more useful than a cropped corner. For PDFs, identifying the relevant page or question helps focus the response.

If information is missing, whacl can ask a follow-up question or hand the conversation to your team. Customers do not need to format every support request perfectly before sending a photo, screenshot, or document.

The practical answer

A WhatsApp AI assistant can understand photos, screenshots, and PDFs when the files are supported and readable. whacl can use those files with your catalog and policies, reply automatically or wait for review, and hand difficult cases to a human with the conversation context intact.

Use automatic replies for clear, repeatable media questions. Use review or handoff for unclear files, private systems, disputes, and exceptions. With WhatsApp, Instagram, and Messenger in one inbox, Claude, GPT, or Gemini available from one setting, and cost-per-reply tracking, you can test that boundary with the files your customers already send.