GOTV — Get Out The Vote — is the most ROI-dense part of any election campaign. In the last 7 days before polling, the campaign is no longer trying to change voters' minds. It's trying to make sure voters who already support the candidate actually show up at the booth. Done well, GOTV adds 1-3% to your supportive turnout. In a close race, that's the entire margin.
This guide is the 14-day pre-polling playbook with AI at the centre. T-14 to T-0, what happens each day, what the AI agent does, what the karyakartas do, how everything synchronises.
The GOTV model: who matters, who doesn't
Before the playbook, understand the GOTV math. Every voter falls into one of four buckets:
| Bucket | Description | GOTV priority |
|---|---|---|
| Strong supporter, consistent voter | Votes every time, supports the candidate | Low — they'll show up regardless |
| Strong supporter, inconsistent voter | Supports the candidate but often doesn't vote | Highest — biggest swing potential |
| Undecided / persuadable | Mixed views, unclear if will vote | Medium — but persuasion window already closed |
| Strong opposition | Supports another candidate | Zero — getting them to the booth helps your opponent |
The entire GOTV apparatus targets the second bucket: supporters who don't reliably vote. This is typically 15-25% of the supportive base — and turning them out is what wins close elections.
The data needed to identify this bucket:
- Voter list with EPIC numbers (from ECI roll)
- Past-cycle turnout history (from ECI's published voter-attendance data, where available)
- Sentiment from the campaign's prior outreach waves (from your AI calls)
Cross-reference: voters who scored Supportive in sentiment but Inconsistent in turnout history. This is the GOTV list.
T-14 to T-7: setup week
The week before polling. The persuasion campaign is winding down; the GOTV apparatus is spinning up.
T-14: GOTV target list assembly
Pull the GOTV list as described above. Typical size for a 50-lakh-voter constituency: 5-15 lakh voters. Smaller is better — focus matters more than coverage at this stage.
Segment the list by:
- Booth (most important — karyakarta-actionable)
- Likely vote time of day (morning vs afternoon vs evening, based on demographics)
- Transport need flag (voters without their own vehicle in rural areas)
- Language / dialect
T-13 to T-10: Wave A GOTV reminder
First reminder wave. Soft introduction: "आपके चुनाव क्षेत्र में मतदान 17 तारीख को है। आपका booth नंबर है X, address Y..."
Tone: informational, not pushy. The agent is preparing the voter, not pressuring.
Volume: full GOTV list, 1 call per voter, ~60-second duration.
Cost: ₹0.50–₹1.00 per call × 5-15 lakh voters = ₹2.5-15 lakh.
T-10 to T-7: ECI Model Code silent period prep
The 48-hour silent period before polling begins at T-2. Anything that would count as persuasion must conclude before then. Schedule:
- Final persuasion wave (if any): T-10 to T-3
- GOTV reminder wave B: T-3 (before silent period)
- Polling-day voter-information calls: T-0 (allowed during voting hours)
Brief the entire team on what's allowed vs banned during silent period. Document it.
T-7: Wave B GOTV reminder
Second reminder wave. More specific: "नमस्ते, मतदान 3 दिन बाद है। आपका booth नंबर X, समय 7am-6pm। क्या आपको transport की ज़रूरत है?"
Voters who answer Yes to transport: ticket created, routed to local karyakarta.
Volume: same list as Wave A, but skip voters who already confirmed via Wave A.
T-3 to T-1: silent period
ECI Model Code prohibits campaign communication in the final 48 hours.
What the AI agent does during this window: nothing. No calls, no SMS, no WhatsApp. The campaign machinery is silent.
What the team does during this window:
- Karyakartas door-knock the GOTV list (this is allowed — it's organic field activity, not bulk communication)
- Polling-booth managers are briefed
- Vehicle dispatch plans finalised
- Local police coordinated for booth coverage
T-0: polling day
The big day. The AI agent comes back online but with strict rule: only voter-information messaging, no persuasion.
6 AM: pre-polling preparation
- Verify all booth assignments are correct (voters have moved, lists may be stale)
- Verify all karyakartas are in position
- Verify the ECI voter-turnout app integration is working
7-9 AM: first reminder wave
Calls to voters in the morning-likely bucket. Reminder: booth number, polling started, what to bring (voter ID).
Volume: ~30% of the GOTV list (morning voters). ~3-5 lakh calls in 2 hours = peak concurrency window.
9 AM-12 PM: ECI turnout monitoring
The ECI's voter-turnout app publishes booth-level turnout numbers every 2 hours. The campaign's data layer ingests this in real time.
For each booth, compute:
- Expected turnout by this hour (from past-cycle baselines)
- Actual turnout from ECI
- Variance — booths under-performing get a second-wave call
12 PM: second reminder wave
Targeted: only the GOTV voters in under-performing booths. Volume: ~1-3 lakh calls. Tone: gentle reminder, "क्या आप अभी तक vote करने नहीं गए? Booth 3 baje तक खुला है।"
3-4 PM: final-hour wave
Last targeted call wave. Voters in still-under-performing booths, who haven't responded to Waves 1 and 2.
Volume: under 50,000 calls. High-touch — these are the voters whose turnout will decide the margin.
4-6 PM: karyakarta closing
The AI has done what it can. The last 2 hours belong to the karyakartas — door-knocking the remaining list, offering rides, escorting reluctant voters to the booth.
The AI's role shifts to support: each karyakarta calls a dedicated number to log their progress, request resources, escalate problems. The agent triages.
6 PM onwards: polling closes
Calls stop. The team awaits results. AI shifts to result-aftermath mode (Wave 4 territory, not this guide).
Coordinating with karyakartas
GOTV doesn't work as AI-only or karyakarta-only. The integration is the value.
What AI does best
- Volume: 5+ lakh reminder calls in a single morning
- Real-time turnout integration: instant identification of under-performing booths
- Multi-lingual coverage: every voter in their preferred language
- Audit trail: every call logged, every voter accounted for
What karyakartas do best
- High-trust last-mile: an elderly voter who won't trust an AI call will trust their long-time party karyakarta
- Physical assistance: helping elderly or disabled voters get to the booth
- On-ground problem solving: incorrect EVM placement, missing voter slip, intimidation incidents
- Booth-level coverage: the karyakarta knows the booth, the neighbours, the local landlords
Integration pattern
The mobile dashboard each karyakarta sees on polling day:
- Their assigned booth(s)
- Today's turnout (vs target)
- List of supportive voters who haven't yet shown up (live-refreshing as ECI data comes in)
- Voters who have requested transport (with phone numbers)
- One-tap status updates ("voter X confirmed at booth", "voter Y refused", "voter Z needs ambulance")
The AI handles the bulk reach; the karyakarta handles the last 50-100 metres of each conversion.
Common GOTV failure modes
1. Calling voters who already voted
Without real-time turnout integration, the AI keeps calling voters who showed up at 9 AM. By the third call, the voter is irritated. Real-time integration is non-negotiable for polling-day operations.
2. Wrong booth information
Voter lists are stale by 6-18 months. A meaningful fraction (3-8%) of voters will have moved booths between roll updates. The AI must verify booth assignment in real time from the ECI roll, not from a campaign-cached copy.
3. Treating GOTV as one wave
A single GOTV wave is fine but suboptimal. The right pattern is 4-5 waves spread across T-7 to T-0, each with a slightly different tone and target. Single-wave GOTV reaches ~60% of the goal; multi-wave reaches 80-90%.
4. Missing the silent period
Campaigns sometimes try to push persuasion content into the final 48 hours. ECI sanctions and reputational damage follow within hours. The silent period is real — respect it.
5. Karyakarta-AI desync
The karyakarta has stale data; the AI has fresh data. They're working at cross-purposes. Synchronisation is a daily problem — solve it with a shared real-time mobile dashboard, not WhatsApp groups.
What GOTV adds to the result
Empirical observations from 2024 cycle deployments where the data is clean:
- AI-driven GOTV adds +1.2% to +2.8% to supportive turnout in targeted segments
- This typically translates to +0.4% to +1.2% to the candidate's total vote share
- In races with margins under 3%, this is the entire margin
- Cost: ~₹20-50 lakh for a 50-lakh-voter constituency's GOTV wave
For a campaign with a tight race, this is among the highest-ROI investments in the entire budget.
Where AiSewak fits
AiSewak's GOTV module ships with:
- ECI voter-turnout app integration (real-time)
- Booth-by-booth target setting and tracking
- Mobile dashboard for karyakartas (Android + iOS)
- Pre-built call wave templates (Wave A, Wave B, polling-day Waves 1/2/3)
- Silent-period auto-pause (calls automatically stop at T-48 hours and resume at T-0 AM)
- Last-mile transport request capture and routing
Where to go next
- The 30-Day Deployment Playbook — getting to launch
- Voter Sentiment Analysis — input data for GOTV targeting
- AI for Booth Workers — karyakarta integration
- The Two-Way Engagement Engine — the architecture underneath
GOTV is where the campaign's entire investment cashes in. The campaigns that build the AI-karyakarta integration before polling day win the close races. The campaigns that don't lose those races by 0.8% margins and never quite understand why.