A 2026 Indian election campaign manager has three direct-contact channels to allocate budget across: voice AI, WhatsApp and SMS. (Old-style IVR robocalls are now technically a fourth, but they're so close to dead that they're not worth a serious budget line.) Each channel has a sharp shape — what it does well, what it does badly, what voters it can and cannot reach.
This guide is the channel mix conversation in numbers. Worth bookmarking before the next budget meeting.
The voter you're trying to reach
Before the channel comparison: a 60-second snapshot of the typical Indian voter the campaign needs to touch.
- Has a phone. ~70% of adults own a personal mobile; ~95% have access to a household phone.
- Phone type varies sharply by demographic. ~55% own a smartphone; ~40% use a basic feature phone. The split correlates strongly with age, income and rural/urban geography.
- Literacy is uneven. ~77% adult literacy nationally, but functional reading (decoding text messages reliably) is lower, especially in older rural voters.
- Language preference is regional and dialectal. A Bihari voter prefers Bhojpuri over Hindi; a Marwari voter prefers Marwari; a Tamil voter cannot be reached in Hindi at all.
The right channel mix has to work for this entire population — not just the urban-smartphone-Hindi-literate cluster that's overrepresented in tech-vendor marketing decks.
Voice AI: what it does
A voice AI campaign call lasts 45–120 seconds. The voter answers, hears the agent introduce itself, gets one or two manifesto-relevant points, and is asked an open question. The agent listens, responds in the right language and dialect, captures sentiment and key issues, and ends.
Reach. Any Indian phone that can receive a call — feature phones, smartphones, dual-SIM, basic-handset, doesn't matter. Effective reach is ~85% of registered voters when the list is clean.
Engagement. 50–70% call completion (voter listens to the agent's full first turn). 30–45% of completed calls produce a meaningful exchange (voter actually says something substantive).
Cost. ₹0.50–₹1.50 per call inclusive of LLM + TTS + telephony, depending on duration and concurrency.
Compliance. Highly auditable. ECI-disclosure rule satisfied by the agent's self-identification line. TRAI-DLT registration mandatory for outbound. DPDP-compliant with proper consent + retention.
Where it wins. Two-way conversation. Rural reach. Senior voter reach. Sentiment capture. Dialect-level personalisation.
Where it loses. Reliance on the voter actually answering the phone. Spam-flag risk if telephony hygiene is poor. Higher per-contact cost than text-based channels.
SMS: what it does
160-character text message sent to a voter's mobile number. No interaction. Standard DLT template, sender-ID, no personalisation beyond a name field.
Reach. Essentially universal — works on every Indian mobile phone going back 20 years. Effective deliverability is 95%+.
Engagement. 5–15% read rate (varies by sender reputation, time of day, content). In practice, most SMS messages are deleted unread.
Cost. ₹0.05–₹0.20 per message — the cheapest direct-contact channel by an order of magnitude.
Compliance. Highly regulated. TRAI DLT registration mandatory for sender + each template. Promotional SMS allowed within DND-respecting windows. ECI Model Code of Conduct restrictions apply during campaign window.
Where it wins. Universal reach. Cost. Time-critical delivery (polling-day reminders, last-mile GOTV).
Where it loses. Engagement. No conversation. No personalisation depth. Effectively invisible to voters who use phones in non-text-dominant ways (a large fraction of rural seniors).
WhatsApp: what it does
Text, image, voice note or video sent to a voter's WhatsApp number via the WhatsApp Business API (official, template-approved) or via informal volunteer-run broadcast networks (unofficial).
Reach. Around 60 crore WhatsApp users in India, against 96+ crore voters. Constituency-level reach typically 50–70% — heavily skewed toward urban, younger, digitally-engaged voters.
Engagement. 20–40% read rate on first delivery; 40–60% on retargeted delivery (the message reappears when the voter opens WhatsApp later). 5–15% of recipients respond with any reply.
Cost. ₹0.10–₹0.40 per templated business message via the official API; effectively zero via volunteer broadcast networks (but legally riskier and capped by Meta's anti-spam rules).
Compliance. Complicated. WhatsApp's own Business API rules require opt-in and pre-approved templates. Political-content templates are scrutinised. Informal volunteer groups operate in a grey zone — they work but attract takedown action and personal-account bans for the volunteers running them.
Where it wins. Rich media (images, voice notes, short videos). Conversation when the voter chooses to respond. Lower per-contact cost than voice AI. Effective for the digitally connected urban-middle-class cluster.
Where it loses. Rural seniors and feature-phone users. Voters who haven't opted in. Legal exposure of informal broadcast networks. Time-of-delivery uncertainty (the voter sees it when they open WhatsApp, not when you send it).
The head-to-head matrix
| Dimension | Voice AI | SMS | |
|---|---|---|---|
| Reach (% of voters) | ~85% | ~98% | 50–70% |
| Engagement (% who consume) | 50–70% | 5–15% | 20–40% |
| Conversation depth | High | None | Medium |
| Cost / contact | ₹0.50–₹1.50 | ₹0.05–₹0.20 | ₹0.10–₹0.40 |
| Cost / meaningful contact | ₹1–₹3 | ₹0.50–₹3 | ₹0.50–₹2 |
| Setup time | 2–3 weeks | 1 week | 3–6 weeks (API) |
| Compliance burden | Medium | High | Variable |
| Personalisation | Very high | Low | Medium |
| Multimedia | Audio | Text only | Text/image/voice/video |
| Senior / rural reach | High | High | Low |
| Smartphone required | No | No | Yes |
The "cost per meaningful contact" row is the one most campaign managers anchor on. When you adjust for engagement quality — what does it actually cost to get one voter to engage with your message — voice AI and SMS are roughly comparable, and WhatsApp is similarly priced for the digitally connected segment.
The right mix, by phase
The wrong question is "which channel". The right question is "which channel for which phase, for which voter segment".
Phase 1: Pre-poll listening (T-180 to T-90)
Primary: Voice AI. Open-ended conversations with a stratified sample produce the inputs to manifesto drafting. WhatsApp surveys work for the digitally connected segment. SMS surveys do not work at all (response rate too low).
Phase 2: Persuasion (T-90 to T-30)
Primary: Voice AI. Two-way conversation is what changes minds. WhatsApp delivers rich media (videos of the candidate, image cards on policies) that reinforces the voice conversation. SMS is a teaser that drives to the voice call ("कल आपको हमारे AI सहायक का call आएगा...").
Phase 3: GOTV (T-7 to T-0)
Primary: SMS for the reminder, voice AI for the conversation. Time-critical SMS hits every voter at the right hour with booth number, polling day, what to bring. Voice AI follows up with the voters who didn't respond to SMS, in the right language, asking if they need help getting to the booth.
Phase 4: Result aftermath (T-0 to T+15)
Primary: WhatsApp for those who voted (thank-you with personalised video), Voice AI for those who didn't (find out why).
Phase 5: Governance (T+15 onward)
Primary: Voice AI inbound. A toll-free or 160-series number that any constituent can call. WhatsApp Business as secondary channel for the digitally connected segment.
Common channel-mix mistakes
Patterns we've seen repeatedly in campaign reviews.
1. WhatsApp-only outreach
Common in first-time candidate campaigns with a small tech-savvy team. Misses 30–50% of the electorate, heavily underrepresents rural and senior voters. Looks productive (the team can see message volume) but doesn't move the dial in competitive booths.
2. SMS-only outreach
Cheap, broad, useless. The engagement is too low to move opinion. SMS is a supporting channel, not a primary one.
3. Voice AI without telephony hygiene
Running voice AI from poorly-managed sender numbers gets the calls flagged as spam by network carriers within 48 hours. The campaign sees declining connect rates and assumes voice AI doesn't work — when actually the telephony was misconfigured. Pre-register on DLT, use a clean Indian sender pool, rotate numbers.
4. Treating channels as competing rather than complementary
The strongest campaigns run all three channels in coordination — voice AI does the heavy persuasion, WhatsApp reinforces with rich media, SMS handles time-critical nudges. Each channel does what it does best.
5. Skipping the dialect tuning
Hindi-only voice AI in a Marwari district produces ~30% completion rate. The same campaign with Marwari-tuned voice AI hits ~65%. The "voice AI doesn't work in our state" finding is almost always a dialect-tuning failure.
Where AiSewak fits
AiSewak ships all three channels in one platform — voice AI as the primary engagement layer, with WhatsApp Business API and DLT-registered SMS as supporting channels. Cross-channel handoff is automatic: a voter who doesn't answer the voice call gets a WhatsApp follow-up; a voter who reads the WhatsApp message but doesn't reply gets a voice call the next day.
The key design choice: a single conversation history per voter across all three channels. When the voice AI calls a voter who has already received three WhatsApp messages, it knows what was said and adapts the opening accordingly. Most multi-channel campaigns lose this stitching and end up sending contradictory or duplicated messages to the same voter.
Where to go next
- Voice AI in Political Campaigns: Technical Guide — the stack underneath
- AI Election Agent Pricing — pricing models per channel
- Hindi-First AI Voice Agents — why the language layer matters
- The 30-Day AI Election Deployment Playbook — execution
The campaigns that win in 2027 and 2029 will not be the ones with the most channels. They will be the ones who allocated the right channel to the right voter at the right phase — and right now the channel doing the heavy lifting is voice AI.