Is AI calling RERA-compliant for Indian real estate developers?

Yes, when architected correctly. RERA does not prohibit any specific calling technology — it regulates representation, disclosure, and accuracy. AI calling is in fact structurally better-suited to RERA compliance than human telecalling because the script is pre-reviewed, immutable, project-tagged, and every call is recorded and auditable. The compliance work is in script review, DLT registration, recording retention, and DND scrubbing — not in the AI itself.

What disclosures must a real estate AI call include under RERA?

At minimum: the developer's RERA-registered legal entity name, the project's RERA registration number for any project being discussed, a recording notice, and (best practice for 2026) an AI disclosure that the caller is an AI assistant. No claim on the call may differ from the project's RERA filing — possession date, carpet area, amenities, project status, pricing must all match the registration certificate.

Are calls to leads from 99acres or MagicBricks promotional or transactional?

Grey area, and safest practice is to treat them as promotional with full DND scrubbing. The buyer initiated contact with the portal, not with the specific developer. The portal monetises the lead to multiple developers. From a TRAI perspective and from a buyer's reasonable expectation, this is not the same as a buyer who filled a form on the developer's own website with explicit consent for that developer's calls. DND scrub these leads.

How long must real estate call recordings be retained for RERA audit?

There is no single statutory minimum across all state RERAs, but practical guidance is 90 days minimum, three years recommended, and seven years conservative. RERA complaints can be filed years after a sale, so the longer retention window protects the developer in any subsequent dispute. Recordings should be encrypted at rest, indexed by project and buyer phone, and retrievable within 24 hours of an audit request.

Can the same AI calling campaign cover multiple RERA-registered projects?

It can, but it should not use a single shared script. Each project must have its own script reviewed against its own RERA certificate, with its own RERA registration number disclosed, its own possession date, its own carpet area schedule, and its own amenity list. The campaign infrastructure can be shared; the scripts must be project-tagged and isolated. This is exactly how Prateek Group runs separate scripts for Prateek Canary, Prateek Edifice, Prateek Grand City, and other projects under one operational deployment.

What is the penalty for RERA non-compliance on a sales call?

Under Section 12 of the RERA Act, misrepresentation that induces a buyer to invest can require the developer to refund the investment with interest and compensation. Under Section 18, failure to deliver as promised triggers refund-with-interest obligations. Under Section 61, the regulator can impose penalties up to 5% of the project cost for general violations, and Section 60 / 12 read with state rules can extend up to 10% in cases of registration-related misrepresentation. A call-based misrepresentation is treated the same as any other; the medium does not exempt the developer.

How does Prateek Group's AI calling stay RERA-compliant?

Prateek Group runs project-tagged scripts on Caller Digital, each reviewed against the specific project's RERA registration certificate before deployment. Every call opens with the registered entity name and the project's RERA registration number. The AI does not deviate from the approved script. Recording retention is built in and indexed by project and buyer for audit retrieval. When a project's RERA filing is amended, the script is updated same-day with the old version archived. Combined with same-day lead activation (2-3 minutes from inbound), this delivered 3x more qualified home buyers reaching their sales team without adding a single telecaller — the practical proof that RERA compliance and high-velocity AI calling are not in tension when the architecture is right.

RERA-Compliant AI Calling for Real Estate India 2026 | Caller Digital

RERA changed Indian real estate sales forever, but most developers are still treating it like a paperwork exercise. They obsess over the brochure language, the website disclosures, the printed prospectus that goes into the buyer's folder at the experience centre. Meanwhile the highest-volume, highest-risk customer-facing surface in their entire business — the outbound call — gets handed to a roster of telecallers reading from a tattered script that nobody has matched against the project's RERA registration certificate in eighteen months.

This is the AI calling angle that real estate is missing, and it is more important than the brochure language they obsess over. A misrepresentation on a printed brochure gets caught in QC. A misrepresentation on a sales call gets recorded in an investigation only after a buyer files a RERA complaint, by which point the developer is staring at a 10% of project cost penalty under Section 12 and 18 of the RERA Act for misrepresentation and false promises.

Layer DPDP 2023 on top — buyer personal data now requires explicit, granular, withdrawable consent — and TRAI's TCCCPR 2018 framework underneath, with its DLT registration and DND scrubbing obligations, and you have a four-layer compliance stack that almost no Indian real estate developer has deliberately architected. They have a CRM, they have a tele-calling team, they have a RERA certificate framed in the experience centre lobby, and they have hope.

Hope is not a compliance strategy. This guide walks through the actual architecture: what RERA requires on a call, where DPDP and TRAI bolt on, the six-step deployment we run for Indian developers, and the practical case of Prateek Group — one of North India's largest developers — running this stack live across Noida and Ghaziabad with a 3x lift in qualified buyers reaching their sales team.

The four-layer compliance stack for real estate AI calling

Most developers have one layer in their head — RERA. The actual stack is four layers, and you fail any one of them at significant penalty.

Layer 1 — RERA (the Real Estate Regulation Act, 2016). Every state has its own RERA authority — UP RERA, Haryana RERA, MahaRERA, K-RERA — but the substantive obligations on outbound communication are aligned. Every call discussing a project must disclose the project's RERA registration number. The developer entity named on the call must be the entity registered with RERA, not a marketing brand or a sister company. No claim made on the call may differ from what is on the RERA filing — possession date, carpet area, amenities, project status (under construction vs ready-to-move), legal title, encumbrances. Misrepresentation on a call is misrepresentation; the medium does not exempt the developer.

Layer 2 — DPDP Act 2023. Buyer phone numbers, names, budgets, family details, financial qualification information are personal data. Section 6 of the DPDP Act requires explicit consent for marketing communication. Section 7 carves out legitimate uses — including service calls about an existing transaction the data principal is party to. A site visit confirmation call is Section 7. A cold call to a scraped database is Section 6 and requires opt-in consent that almost no developer can actually evidence.

Layer 3 — TRAI TCCCPR 2018. DLT (Distributed Ledger Technology) registration of every script template before it goes on a wire. DND (Do Not Disturb) preference scrubbing for any promotional call. Promotional vs transactional classification — and the right number series (140x for promotional, 1600 for transactional/service). The 9 AM to 9 PM calling window for promotional. Penalties for breach include disconnection of the calling line and reputational disclosure to TRAI's central database.

Layer 4 — State-specific consumer protection. Several states have additional consumer protection rules — Maharashtra's MOFA (now subsumed by RERA but still cited), state-level real estate ombudsman provisions, and CCPA's evolving guidelines on misleading advertisement that explicitly cover oral representation.

Skip a layer and the regulator on that layer is the one who finds you, not the others.

What RERA actually requires on a call

The substantive RERA obligations on a sales call are narrower than developers fear and more specific than they assume. Get these right and the rest is form-filling.

The developer's RERA-registered legal entity name must be stated. Not "Acme Realty" if RERA registration is "Acme Realty Private Limited." Not the marketing umbrella brand if individual SPVs hold each project. The buyer must hear the name they will see on the agreement to sale.

The project's RERA registration number must be disclosed for any project being discussed. UP RERA numbers look like UPRERAPRJ123456. MahaRERA like P51800012345. The format varies; the disclosure obligation does not. If the AI is qualifying a buyer for Prateek Canary, the AI says the Prateek Canary RERA number on that call.

No possession date claim that contradicts RERA filings. If RERA filing says possession December 2027, the AI does not say "we'll hand over by mid-2027" because the sales team thinks construction is ahead. The script says December 2027. Period.

No carpet area number that differs from RERA registration. Carpet area is the RERA-defined number. Super built-up is a marketing number. The AI states carpet area when asked area; super built-up only with explicit framing.

No amenities mentioned that are not on the RERA filing. If the gym and clubhouse are filed but the rooftop infinity pool was a CGI render that didn't make the final filing, the AI does not mention the pool. Ever. This is the failure mode that triggers the most RERA complaints.

No misrepresentation of project status. "Under construction" if under construction, "ready-to-move" if OC has been received, "RERA-approved" only if registration is granted (not "applied").

The script for sales calls must be reviewable against the project's RERA certificate before going live. This is the architectural requirement most vendors duck. You need a paper trail showing each line of the script was checked against each clause of the registration before the script went into production.

Recording retention long enough for audit. RERA complaints can be filed for years after a sale. 90 days is a floor. Three years is recommended. Seven years is conservative.

Transactional vs promotional — the distinction that matters most

The single most important compliance decision in real estate AI calling is the transactional vs promotional classification of each call type, because this determines whether DND scrubbing and the 9-9 window apply.

Site visit confirmation for an interested buyer who has provided their number — service call, transactional, no DND scrubbing required. The buyer initiated the relationship; you are confirming a logistical detail.

Cold call to a leads database from a portal scrape — promotional, full DND scrubbing mandatory. The "scrape" word matters. If you bought the database, scraped it, or aggregated it without explicit consent for your developer's calls, you are promotional.

"New launch" announcement to a past leads database — promotional. The buyer expressed interest in a different project two years ago. They did not consent to perpetual marketing.

Inquiry follow-up where the buyer initiated contact within the last 90 days — service-related, lighter compliance, but still careful. The 90-day window is not codified in TRAI; it is a practical risk-management heuristic that aligns with how regulators interpret "active inquiry."

The most common mistake real estate developers make is treating all leads from 99acres, MagicBricks, and Housing.com as inquiries that bypass DND. They do not. The buyer initiated contact with the portal, not with the developer specifically. The portal sells the lead to multiple developers. From the buyer's perspective, they did not opt into ten developer call centres calling them. From a TRAI perspective, this is the grey area where most penalty actions land. Safest practice is DND scrubbing on portal-sourced leads. (See our TRAI DND compliance guide for the full treatment.)

The six-step RERA-compliant AI calling deployment

Here is the actual architecture we use with Indian developers. Six steps, run sequentially, no skipping.

Step 1 — Inventory of all RERA-registered projects with their certificates. Pull every certificate, every amendment, every quarterly progress report. Tabulate the four critical fields: legal entity, RERA number, possession date, carpet area schedule. If a project has been amended (extension of completion date, change of amenities), use the latest amendment. This is project ops housekeeping that should already exist; in practice it rarely does in clean form.

Step 2 — Script writing with project-by-project compliance review. One script per project. The temptation is to write one master script with project-name variables. Resist it. Possession dates differ. Amenities differ. RERA numbers differ. Pricing per sq ft differs. A master template invites cross-project contamination — the failure mode where a telecaller (or an AI fed bad context) mentions Project A's pool while qualifying for Project B. Each script gets reviewed against that project's certificate by someone who can be held accountable — typically the legal/compliance head, not the marketing head.

Step 3 — Mandatory disclosure block in every script. First fifteen seconds of every call: registered entity name, project RERA ID, recording notice, AI disclosure ("I'm an AI assistant from..."). The AI disclosure is increasingly important — Indian regulators have signalled that undisclosed AI in consumer-facing voice will face scrutiny under unfair trade practice rules in 2026. Disclose upfront. It does not reduce qualification rates; it raises trust.

Step 4 — DLT template registration. One template per project per call type. Lead qualification template for Prateek Canary is a different DLT template from site visit confirmation for Prateek Canary. And both are different from the equivalent templates for Prateek Edifice. Yes, this is more registrations. No, the consolidation shortcut is not worth the audit risk.

Step 5 — NDND scrubbing for promotional campaigns. Live integration with the National Customer Preference Register, scrubbed at call dispatch time, not at campaign upload time. Buyer preferences change daily; scrubbing a list at upload and dialling three days later is not compliant.

Step 6 — Recording retention infrastructure. Encrypted at rest, indexed by project + buyer phone + date, retrievable in under 24 hours for RERA complaint response. 90-day minimum, three-year recommended. This is the layer that turns AI calling from a compliance liability into a compliance asset — every call is recorded, indexed, and reviewable, which is something the human telecalling channel almost never delivers cleanly.

For the broader DPDP architecture this sits inside, see our DPDP compliance for AI calling guide.

Case study: how Prateek Group runs RERA-compliant AI calling at scale

Prateek Group is one of North India's leading real estate developers, with landmark projects across the Noida and Ghaziabad belt — Prateek Grand City, Prateek Edifice, Prateek Canary, Prateek Wisteria. Multi-tower communities, premium positioning, the kind of inventory that draws thousands of inquiries a month off 99acres, MagicBricks, Housing.com, the developer's own website, Meta lead forms, and walk-in registrations at experience centres.

Their challenge before deploying voice AI was the one every multi-project developer in NCR will recognise. Lead volume was not the problem; lead conversion was. Inquiries arrived twenty-four hours a day. The sales team was buried in unqualified contacts — buyers asking about projects out of their budget, NRIs with timeline mismatches, brokers fishing for inventory, and a hard-to-find core of genuinely qualified home buyers whose interest cooled within hours of the form fill.

The industry data on lead response time is brutal. A lead contacted within five minutes of submission is qualified at multiples of the rate of a lead contacted at the one-hour mark. By two hours, you are talking to someone who has already been called by three of your competitors. Prateek's sales team — even running fully staffed across two shifts — was not winning the speed-to-lead battle. They were winning the conversion battle on the leads they could actually reach in time, which is why senior leadership did not want to dilute that team with high-volume cold qualification.

The deployment with Caller Digital was structured as an AI lead qualification and site visit booking layer that activates within 2–3 minutes of a lead arriving from any source — portal, website, Meta. Voice AI calls the lead in Hinglish, qualifies on budget band, configuration preference (2BHK, 3BHK, 4BHK), location preference within the NCR cluster, and timeline (ready-to-move vs under-construction comfort). For qualifying leads, it books a site visit slot directly into the sales calendar. For non-qualifying leads, it captures the disqualification reason and routes accordingly.

The result Prateek Group reports: 3x more qualified home buyers reaching the sales team, without adding a single telecaller.

The compliance angle is the part most developers do not appreciate when they evaluate AI calling. Every script Prateek runs is reviewed against the specific project's RERA registration certificate before going live. The AI does not deviate from the approved language — there is no "let me check with my supervisor and call you back with a different possession date" failure mode. Every call opens with the registered entity name and the project's RERA registration number. Recording retention is built into the platform, indexed by project and buyer phone, ready for audit. When the RERA filing for any project is amended (and amendments do happen — completion extensions, amenity revisions), the script for that project is updated within the same business day and the old version is archived with timestamps.

This is one of the under-appreciated structural advantages of AI calling versus human telecallers in a RERA world. Human telecallers improvise. They build rapport with the buyer by saying "yeah, the pool will probably be ready before possession" because they want to close the slot. They mix up Project A's amenities with Project B's because they are juggling four campaigns. They forget the RERA disclosure on call seven of the day because nobody is auditing call seven. The AI does not improvise. It does not get tired. It does not optimise for closing a single slot at the cost of compliance. Every call is on script, every script is on the RERA certificate, and every recording is in the audit log.

For the lead qualification flow we run with Prateek and other developers, see /use-cases/lead-qualification-follow-up and /use-cases/appointment-booking-reminders.

The five most common RERA compliance failures in real estate calling

After reviewing hundreds of telecalling recordings during compliance audits, the failure modes are remarkably consistent. Five patterns explain almost every RERA complaint that originates from a phone conversation.

Possession date drift. The RERA filing says December 2027. The site engineer tells the sales head "we're tracking ahead, probably September." That trickles into the telecaller's head as "around mid-2027." That trickles to the buyer as "you'll get it next year." The buyer hears a commitment. The RERA filing has the actual commitment. When the project hands over in February 2028, the buyer files a complaint citing the call.

Amenities promised that are not on the RERA filing. Marketing renderings show a jogging track, a meditation pavilion, a co-working lounge. The RERA filing was made before those amenities were finalised and only includes the gym, clubhouse, and pool. Telecallers describe the renderings. Buyers expect what they were told. The complaint follows.

Carpet area inflation. The RERA-mandated metric is carpet area. The marketing-friendly metric is super built-up. A telecaller asked "how big is the 3BHK?" who responds with the super built-up number without explicit framing has misrepresented the carpet area, full stop.

Misrepresenting project status. "RERA-approved" said when the registration is only filed and pending. "Under construction" said when the OC has not been received but possession is being offered. "Ready to move" said when fit-and-finish is incomplete. Each of these is a Section 12/18 misrepresentation.

Cross-project contamination. A telecaller running campaigns for four projects mentions the wrong project's specs to the wrong buyer. Different RERA number, different possession date, different carpet area, different price band — all swapped.

AI calling solves all five structurally. The script is pre-written, pre-reviewed, project-tagged, and immutable mid-call. The AI does not improvise possession dates, does not invent amenities, does not confuse super built-up with carpet area, does not misclassify project status, and cannot mix up two projects because the campaign metadata locks the script to the project ID. Every call is recorded and auditable. The audit trail itself is the defence in a RERA complaint hearing.

Eight questions to ask your AI calling vendor about RERA compliance

If you are evaluating an AI calling vendor for real estate, these are the eight questions that separate a serious platform from a generic one. (For a broader vendor comparison, see our 2026 platform buyer's guide.)

Can you produce the script for any specific project on demand? A vendor who cannot retrieve the live script for Prateek Canary in under five minutes does not have project-tagged campaigns.
Is the script reviewable against the project's RERA certificate? Ask for the format. Side-by-side line review is the standard. A PDF dump is not.
Does the AI deviate from script under any circumstance? The answer should be no. If the answer is "the AI uses contextual judgement," that is a polite way of saying improvisation, which is exactly the failure mode RERA penalises.
How are recordings retained and for how long? Encrypted, indexed, and three years minimum.
What is your audit response timeline if a RERA complaint is filed? Under 24 hours to retrieve a specific call recording is the bar.
Can you handle separate scripts for separate projects in the same campaign? Multi-project developers need this. A vendor that runs one campaign with one script across projects is a compliance accident waiting to file itself.
Do you maintain a paper trail of script approval against RERA filings? Versioned, timestamped, with named approver. This is what a regulator asks for.
What happens if the RERA filing changes mid-campaign? Same-day script update with old version archived is the answer you want.

The Hindi script challenge in North India real estate

There is a craft layer beneath the compliance layer that almost every global voice AI platform fails. North India real estate sales — Delhi, Noida, Gurgaon, Ghaziabad, Faridabad — operates predominantly in Hinglish. The buyer expects Hindi for warmth and rapport ("haan ji, aap kahaan se baat kar rahe hain") and English for the transactional vocabulary that has no clean Hindi equivalent ("carpet area 1,250 sq ft, possession December 2027, RERA number UPRERAPRJ-bracket-bracket-bracket").

The script writing challenge is non-trivial: Hindi sentence structure with English transactional terminology, code-switched naturally at the right points, with the regulatory disclosures (RERA number, recording notice, AI disclosure) read crisply in English because those terms are the legally registered terms.

Caller Digital's models are trained specifically on Indian telephony Hinglish — the actual conversational patterns of Indian buyers on Indian phone networks, not a Hindi model retrofitted with English code-switching. This is one of the structural reasons Prateek Group's deployment works at scale across the NCR buyer base, and it is the gap most global voice AI platforms cannot close. (Our Hinglish code-switching guide goes deeper.)

What good RERA-compliant AI calling looks like in 2026

The opportunity for Indian developers in 2026 is sharper than the compliance overhead suggests. The developers who deploy RERA-compliant AI calling first will out-qualify their competitors by 2-3x on every leads source — portal, website, Meta, walk-in — because they will reach buyers in the two-minute window where intent is highest, with scripts that are tighter and more compliant than any human telecaller can sustain across an eight-hour shift.

The compliance overhead is real but not difficult once architected correctly. Four-layer stack, six-step deployment, project-by-project script review, three-year recording retention. That is the playbook. Prateek Group is the proof case — running it live across multiple projects in Noida and Ghaziabad, qualifying 3x more home buyers without adding a single telecaller, with every call disclosed, scripted, recorded, and audit-ready.

The developers who do not deploy this in 2026 will spend 2027 explaining to RERA hearings why their telecalling team made commitments their projects cannot keep. The developers who do will spend 2027 closing the buyers their competitors lost in the cold-lead window.

To explore what this looks like for your projects, see our real estate industry page and the AI caller India overview, or look through case studies of comparable developer deployments.

RERA-Compliant AI Calling for Indian Real Estate Developers: The 2026 Field Guide