From 1 October 2026, replying to your own customers on WhatsApp costs money. Free-form messages sent inside the 24-hour customer service window — the plain text reply, the image, the PDF brochure, the set of quick-reply buttons — are billable again after a monthly free allowance. They had been free since November 2024.
The per-message rate is small. The behaviour it exposes is not. Most WhatsApp deployments we audit in India send three to five messages where one would do, and that habit was free until last week. Now it is a line item.
What actually changed
Four things are worth getting exactly right, because the summaries floating around get at least one of them wrong.
First, the free allowance is 1,000 delivered service messages per business phone number, per calendar month. Per number, not per WhatsApp Business Account. If you run three numbers, you have three separate pools of 1,000. The allowance does not roll over and resets on the 1st.
Second, you are billed on delivered messages, not sent ones. A message Meta never delivers is not charged.
Third, service messages are billed at your market's utility rate. In India that is ₹0.115 per message at the time of writing, which with 18 percent GST lands at roughly ₹0.136 — the ₹0.14 figure you have probably seen quoted is this number rounded up.
Fourth, and this is the one that catches teams out: it does not matter whether a human agent or a bot sent the reply. An automated flow that fires four bubbles in sequence is four billable messages, not one interaction.
Messages your customers send to you remain free. You are only ever charged for what you send.
The maths, in rupees
For a small operation this barely registers. A presales desk on one number handling 8,000 service messages a month pays for 7,000 of them — about ₹950. That is not a budget conversation.
At scale it changes shape. Take a support or presales team pushing 100,000 service messages a month from a single number. That is 99,000 billable messages, roughly ₹13,400 a month, about ₹1.6 lakh a year, for replies that cost nothing in September.
Now apply the multiplier almost nobody has measured. If your bot answers each enquiry with four separate bubbles — a greeting, the answer, a follow-up question, a menu — consolidating those into one message takes the same 25,000 conversations from 100,000 billable messages to 25,000. After the free allowance that is roughly ₹3,300 a month instead of ₹13,400. You have cut about ₹1.2 lakh a year by editing message templates, with no change to what the customer learns.
That is the honest headline. The rate is not the problem. Message sprawl is the problem, and the rate simply made it visible.
Where the waste usually hides
When we audit a WhatsApp flow, the billable padding is almost always in the same five places.
Acknowledgement bubbles. "Thanks for reaching out!" followed a second later by the actual answer. Two delivered messages, one unit of information. Fold the acknowledgement into the answer.
Drip-fed bot menus. A greeting, then a question, then the options, each as its own send because it read better in the flow builder. One message with the options attached does the same job.
Typing-indicator theatre. Some teams send filler messages to simulate a human pause. That is now money spent to look slower.
Re-asking what you already know. If the lead came from a Meta ad with their name and project interest attached, asking for the name again costs a message and annoys the buyer.
Conversations that should have been calls. A twelve-message back-and-forth to pin down budget, configuration, and a site-visit slot is twelve billable messages and a slow customer experience. The same qualification on a voice call is one conversation.
What to do about it this month
In rough order of return on effort.
Consolidate your replies. Go through your automated flows and count the delivered messages per customer turn. Anything above one needs a reason. This is the single biggest lever and you can ship it this week.
Resolve in fewer turns. Answer the question the customer is actually going to ask next. A reply that pre-empts the obvious follow-up removes two messages, not one.
Use free entry points properly. Conversations opened from Click-to-WhatsApp ads and Facebook Page call-to-action buttons still carry a free window. If your team replies after it closes, you have paid for a conversation that was free. Check the current window length in your WhatsApp Manager — the published figure has moved recently, and it differs by entry point.
Check your provider's markup. Meta's rate is the same through every BSP. Anything above it is your provider's margin, and it is negotiable.
Separate numbers by use case. Running bulk broadcast on a different number from nurturing was already the right call for quality-rating reasons we wrote about here, and each number now carries its own 1,000-message allowance. Treat that as a side benefit of a sound setup, not a reason to spin up numbers you have no genuine use for — Meta is reasonably good at spotting allowance farming, and a damaged quality rating costs far more than the messages you saved.
The structural fix: stop using chat for qualification
Everything above trims the bill. The bigger move is to stop paying per message for work that a phone call does in one pass.
Qualification is a conversation with branches. Budget, configuration, timeline, location preference, financing — in chat that is a dozen turns, each one billable, spread over hours because the customer replies when they feel like it. On a call it is three minutes, and you get tone, hesitation, and intent that chat never surfaces.
The pattern that works, and the one we build for, is voice first and WhatsApp for what chat is genuinely best at. An AI voice agent calls the lead within 60 seconds in Telugu, Hindi, or English, qualifies them properly, and books the slot. The WhatsApp agent then sends one message: the confirmation, the location pin, the brochure. One delivered message instead of twelve, and a better experience at the same time.
That inverts the economics of the pricing change. Teams running chat-first qualification just watched their cost per lead rise. Teams running voice-first qualification with WhatsApp as the confirmation layer barely feel it, because they were never sending twelve messages per lead.
Two things to verify in your own account
We would rather flag uncertainty than have you plan around something that turns out to be wrong for your account.
Several providers report that utility templates sent inside the service window also lost their free status on 30 September 2026, with no free tier at all. Meta's own notice that we have seen covers service messages specifically. If utility templates are a meaningful share of your volume, confirm the current treatment in your WhatsApp Manager before you model it.
Second, if you are a government body or a registered non-profit, service messages stay free beyond the allowance through 31 December 2027. Worth checking your eligibility rather than assuming the standard rate applies.
The short version
₹0.136 a message is not what should worry you. Sending four messages to deliver one answer, for every lead, forever, is what should worry you — and until 1 October there was no reason to notice.
Audit your flows for message count per customer turn, and move qualification off chat and onto a call. If you want a second pair of eyes on where your WhatsApp volume is actually going, talk to us — we do this audit for businesses across Hyderabad and India, and the message-count problem is usually visible within an hour.