Contents
01
Summary
Before a buyer speaks to anyone, two things happen: they ask a question and read the answer, then they land on a project page. We measured both across four cities and found buyers slipping away at each, before a salesperson sees the enquiry.
- AI answers send buyers to portals. Portals are 38.9% of every source the AI engines cited. Developers’ own sites are 12.2%. A buyer who follows the sources is far more likely to land on a portal than on your site.
- Being named is not being cited. 58 of the 60 developers in our sample were named in answers about their city, 771 times in all. Their own sites were cited 103 times.
- Project pages are slow on a phone. 55 of 57 pages took more than 4 seconds to show their main content in a lab test on a slow mobile connection. Google calls anything over 4 seconds poor.
- Speed to lead in India is unmeasured. We found no credible public study of how fast Indian developers answer an enquiry. In the US, a 2011 audit of 2,241 companies found 23% never replied to a web lead[1], and a 2025 mystery shop of 50 home builders found 42% never phoned[2].
- The rules now reward speed. TRAI’s 2026 amendment lets a business contact a buyer on the strength of an enquiry for seven days only[11]. A slow first reply now spends time the rules give only once.
02
When buyers ask AI, it sends them to portals
We asked three AI engines 120 questions a home buyer might ask, 30 per city, such as “Which are good 2 BHK projects under ₹90 lakh in Whitefield?”. None of the questions named a developer. We then looked at every source the 360 answers cited.
Portals take the largest share of citations
The answers cited 2,225 sources across 767 websites, about 6.18 per answer. Property portals were the largest group by a wide margin. Developers’ own sites came third, behind news and media.
Counted by answer rather than by citation, the gap is the same. 87.8% of answers cite at least one portal. 55.6% cite a portal and no developer site at all. Only 4.7% cite a developer’s site with no portal beside it.
Which portals the buyer meets
Housing.com and Square Yards were each named or cited in more than a third of answers. Each engine has its favourites: Housing.com appeared in 76.7% of Perplexity’s answers but in 0.8% of Claude’s.
The most cited websites were all portals or listings sites, with one broker site (66 answers) and one business newspaper in the top ten. The best cited developer site appeared in 21 answers.
| Website | Type | Answers citing it (of 360) |
|---|---|---|
| housing.com | Portal | 138 |
| squareyards.com | Portal | 133 |
| magicbricks.com | Portal | 72 |
| nobroker.in | Portal | 63 |
| justdial.com | Portal | 41 |
| realestateindia.com | Portal | 40 |
| 99acres.com | Portal | 36 |
| ghar.tv | Portal | 34 |
| economictimes.indiatimes.com | Media | 28 |
Named is not cited
Two thirds of answers (66.7%) name at least one developer. So we took a sample of 60 developers, 15 per city, and checked two things in the answers about their own city: was the developer named, and was its own website cited as a source?
Large listed developers are usually cited when they are named. Smaller developers are not: a small regional developer was named in a median of 3 of the 90 answers about its city, and 14 of 20 never had their own site cited. When AI does mention them, the buyer is sent somewhere else to learn more.
By city
Each city’s answers named between 138 and 152 different developers, so no single name owns a city. Bengaluru is the most concentrated: its most named developer appeared in 47.8% of answers. Pune leans hardest on portals, Bengaluru on project microsites, and Hyderabad on news sites.
| Share of answers that… | Bengaluru | Hyderabad | Pune | Mumbai (MMR) |
|---|---|---|---|---|
| Cite a portal | 87.8% | 86.7% | 93.3% | 83.3% |
| Cite a developer’s own site | 42.2% | 37.8% | 33.3% | 34.4% |
| Cite a project microsite | 46.7% | 35.6% | 13.3% | 22.2% |
| Cite a broker or adviser | 50.0% | 34.4% | 44.4% | 58.9% |
| Cite news or media | 32.2% | 44.4% | 30.0% | 36.7% |
| Name a developer | 72.2% | 64.4% | 64.4% | 65.6% |
| City | Developers named | Most named developer, share of answers | Developers named in 10%+ of answers | Microsites, share of citations | Media, share of citations |
|---|---|---|---|---|---|
| Bengaluru | 138 | 47.8% | 10 | 16.8% | 10.6% |
| Hyderabad | 150 | 31.1% | 6 | 10% | 23.5% |
| Pune | 152 | 31.1% | 6 | 3.4% | 12.6% |
| Mumbai (MMR) | 142 | 33.3% | 10 | 6.1% | 17.1% |
By engine
The engines behave differently. Perplexity cites the most sources and nearly always a portal. The ChatGPT-style engine cites the fewest and is least likely to cite a developer’s site. Claude names developers most often.
| Engine | Citations per answer | Cites a portal | Cites a developer site | Names a developer | Portal seen most |
|---|---|---|---|---|---|
| Perplexity | 10.88 | 99.2% | 55.8% | 69.2% | Housing.com 76.7% |
| ChatGPT-style | 3.07 | 84.2% | 21.7% | 56.7% | MagicBricks 45.0% |
| Claude | 4.60 | 80.0% | 33.3% | 74.2% | Square Yards 35.8% |
Caveats: one run on one day, through the engines’ API models rather than the consumer apps, which may answer differently. Developer names were matched by pattern, so a few broad names may be over-counted. The Mumbai questions cover the wider MMR.
03
The project page is where enquiries leak
A buyer who clicks through to the developer lands on a project page. We checked one live project page for each of the same 60 developers: how fast it loads on a phone, and what it asks of a buyer who wants to enquire.
Slow on a phone
We ran Google’s Lighthouse test on each page as a phone on a slow mobile connection[15]. 57 pages could be measured. The median performance score was 43 out of 100, and the range 9 to 74. None reached 90, the score Lighthouse calls good. 36 of 57 scored under 50.
The number a buyer feels is how long the main content takes to appear (the largest contentful paint). Google calls 2.5 seconds or less good and anything over 4 seconds poor[16]. The median page took 9.9 seconds. 55 of 57 took more than 4 seconds; one met the 2.5 second mark. The median page weighed 6.4 MB.
Size does not buy speed. The median score was 44 for large listed developers, 46 for mid-sized and 38 for small regional ones.
What the page asks of a buyer
Every page we could read had an enquiry form. What happens around the form is less consistent. Each share below is of the pages where we could read the answer, so the base changes from row to row.
- Consent. Under a third of forms had no consent box, and the gap varies by city: 91.7% of Mumbai forms had one, Pune 2 of the 9 we could read. One form had its box ticked in advance, and three used a line of text saying the enquiry overrides Do Not Disturb instead of a box. Neither looks like the “clear affirmative action” India’s data law asks for (chapter 6).
- WhatsApp. WhatsApp had more than 400 million users in India by 2019 (chapter 5), yet only 39.7% of pages link to it. Mumbai pages were least likely to (26.7%).
- Phone. Just over half of Mumbai pages showed a number (53.3%). Several Pune and Mumbai pages showed a Bengaluru 080 number, likely a call centre or virtual line.
- RERA. 53 of 57 pages showed a RERA registration number. All four that did not were in Bengaluru or Hyderabad.
Forms were short: a median of 4 fields, from 3 to 8.
| City | Median speed score | Consent box | WhatsApp link | Phone shown | RERA number shown |
|---|---|---|---|---|---|
| Bengaluru | 37 | 85.7% of 14 | 46.7% of 15 | 92.9% of 14 | 86.7% of 15 |
| Hyderabad | 43 | 63.6% of 11 | 35.7% of 14 | 76.9% of 13 | 85.7% of 14 |
| Pune | 38 | 22.2% of 9 | 50.0% of 14 | 92.9% of 14 | 100% of 13 |
| Mumbai (MMR) | 46 | 91.7% of 12 | 26.7% of 15 | 53.3% of 15 | 100% of 15 |
Caveats: 11 pages build their form in the browser, so their form details are unknown and left out of those bases. 6 developers were swapped for others because their sites could not be read, which tilts the sample towards sites that work without JavaScript. Pune’s consent figure rests on 9 readable forms.
04
Speed to lead: what the research says
Everyone in sales agrees a fast reply matters. The published evidence is older and thinner than the quotes suggest, and none of it is from India.
US benchmarkThe 2011 audit of 2,241 companies
Researchers writing in Harvard Business Review sent a web enquiry to 2,241 US companies and timed the reply[1]. 37% answered within an hour. 23% never answered. Among those that replied within 30 days, the average wait was 42 hours.
In a separate study of 1.25 million leads in the same article, firms that tried to contact a lead within an hour were nearly seven times as likely to qualify it (have a meaningful conversation with a decision maker) as firms that tried an hour later, and more than 60 times as likely as firms that waited a day or more.
US benchmarkThe 2025 home builder mystery shop
In 2025, Blue Gypsy, Melinda Brody & Company and Denim Marketing sent online enquiries to 50 US home builders, nearly 30 of them from the Builder Top 200, and tracked their follow-up for 30 days[2]. 42% never made a single phone call. 24% never sent a personal email. None sent a personal email within five minutes.
India: no published number
We looked for a survey or mystery shop of how fast Indian developers or brokers answer an enquiry, and found none we would cite. What circulates are vendor blogs with unsourced figures. Treat any India speed-to-lead statistic you see quoted with care, including the widely shared “contact within 5 minutes” multipliers, which we could not trace to a primary source.
We have written a protocol to measure it the same way for every developer: web form, WhatsApp and phone, at set times of day, over 14 days. It has not been run. We plan it for a later edition.
05
The market you are selling into
Sales in all four cities fell in 2025 and have steadied in 2026, while unsold homes keep rising. With more projects for each buyer to choose from, every enquiry is worth more to the developer who answers it.
ANAROCK counts about 3,95,625 homes sold across India’s seven largest cities in 2025, down 14% from 4,59,645 in 2024. Sales value rose about 6%, to over ₹6 lakh crore[3]. All four of our cities sold fewer homes.
2026 has steadied. In July to September, the seven cities sold about 1,00,200 homes, +3% on a year earlier. But developers launched about 1,14,300 in the same quarter, and supply has outrun sales in every quarter of 2026. Unsold stock reached 6,30,600 homes, up 12%, or about 19 months of sales[5].
| City | Sold, Jul to Sep 2026 | Sales, change on a year | Unsold homes, 30 Sep 2026 | Unsold, change on a year | Months to sell at current pace |
|---|---|---|---|---|---|
| Mumbai (MMR) | 31,750 | +5% | 1,98,050 | +12% | 19 |
| Bengaluru | 16,650 | +12% | 80,200 | +35% | 15 |
| Pune | 15,700 | -5% | 86,900 | +5% | 17 |
| Hyderabad | 12,950 | +15% | 1,14,700 | +20% | 28 |
Hyderabad holds 28 months of stock, the most of the four. Bengaluru’s unsold homes rose 35% in a year. In both cities more projects are competing for each buyer.
Who is buying, and where they are
- Listed developers sold nearly ₹1.95 lakh crore of homes in FY26, up 17%, by a press tally of 28 companies’ disclosures[6].
- India had 1,114.17 million internet subscriptions at the end of June 2026, almost all of them wireless[7]. Subscriptions are not people, but nearly every connection is mobile.
- WhatsApp last confirmed more than 400 million users in India in 2019[8]. Meta has published no newer figure we could find.
- No credible national figure exists for the share of homes bought by NRIs, or for what a lead costs. Figures that circulate come from agencies and vendors using their own client data. Use your own numbers.
06
The new rules on calling buyers
India’s rules on calls and data changed in 2025 and 2026. They favour developers who reply fast, keep proof of consent and call from registered numbers. This is a plain reading of the primary texts, not legal advice.
An enquiry opens a seven-day window
TRAI’s Third Amendment to its commercial communication rules, announced on 18 September 2026, lets a business send commercial communications on the strength of a customer’s enquiry for seven days only. The enquiry must be made in writing or digitally, and the business must keep it in a form that can be checked[11].
Promotional calls land between 10:00 and 21:00
Since the 2018 regulations, the bands before 10:00 and after 21:00 are off by default for every customer, unless they switch one on[9]. “Real Estate” is its own category a customer can block, by sending BLOCK 2 to 1909.
Unregistered numbers put every line at risk
Under the 2025 Second Amendment, any commercial call from a sender not registered with an operator counts as unsolicited[10]. For such a sender, five complaints from different people within ten days lead to suspension of the sender’s outgoing lines and an investigation. A first proven violation bars all outgoing services, including SIP trunks and SIM cards, across all operators for 15 days. A repeat means a year’s disconnection and blacklisting. The 2026 amendment lowers the trigger to three complaints when the operator’s own systems have also flagged the number[11].
AI and auto-dialled calls must be declared
The 2026 amendment defines calls started by software, including robo-calls and artificial voices, and requires every business using them to declare them and the numbers used to its operator in advance. Undeclared ones are treated as unsolicited[11].
Consent under the data law
- Consent must be free, specific, informed, unconditional and unambiguous, given by a clear affirmative action, and come with a notice of what data is used and why. Withdrawing it must be as easy as giving it[12].
- Data a person gives voluntarily, for a purpose, may be used for that purpose. The Act’s own example is a buyer messaging a broker to find a flat. Once the buyer says they no longer need help, the broker must stop.
- Penalties reach ₹250 crore for failing to keep data safe and ₹50 crore for other breaches. Most day-to-day duties under the Rules, including the notice, start around 13 May 2027[13].
- WhatsApp’s own policy allows messages only to people who gave their number and opted in, and requires honouring any opt-out, on WhatsApp or off it[14].
Old leads are the hard case. A lead from last year has no live enquiry window. Legacy consents count under the 2026 amendment only where they were obtained by verifiable means and registered on the operators’ platform[11].
07
The playbook
Six practices follow from the findings. None needs a new team; most need a change in what the existing team measures.
- Be found where AI looks. Keep each project’s facts (price range, configurations, location, possession date, RERA number) on its own page, in plain text AI engines can read, and keep portal listings complete and current. The engines we tested cite pages that state these facts; make your project page one of them.
- Make the project page fast on a phone. Aim for the main content in under 2.5 seconds. Compress images, load video only when asked, and put the enquiry form near the top. Test with Google’s field data, not only the lab.
- Ask for consent the right way. An unticked box with plain words naming your company and the channels you will use, a short notice, and a stored record of what the buyer agreed to and when. No pre-ticked boxes, no “overrides DND” lines.
- Answer within minutes, day and night. An enquiry can arrive at any hour, and the seven-day window starts at once. Reply on the buyer’s channel with an answer, not a holding message, then book the visit.
- Re-engage old leads with consent, not cold calls. Ask for fresh consent by a registered route before calling an old lead, and stop at the first no. Calling from unregistered 10-digit numbers risks every line the company has.
- Tag every source. Put a source tag on every ad, portal and QR code, and carry it into the CRM, so each rupee of spend traces to enquiries, site visits and bookings.
08
Where Intend fits
Intend, which published this report, is software for real estate developers and agents, built and run by Kaigen Labs. It answers buyer enquiries within seconds on WhatsApp, phone calls and the website, from the developer’s own project information, books site visits, and follows up. Before any AI call it checks consent, calling hours, registration and opt-outs, and it applies the seven-day enquiry window. It joins ad spend to the CRM to show what each campaign costs per enquiry, site visit and booking.
That covers practices 4 to 6 above. Practices 1 to 3 are work on your own website and listings, and no software does them for you. To see Intend on your own projects, read how it works for developers or book a demo.
09
Method and sources
AI search
- Run on 26 October 2026, 21:25 to 21:38 UTC. 120 buyer questions, 30 per city (Bengaluru, Hyderabad, Pune, Mumbai MMR), written to be neutral: none names a developer.
- Three engines through their APIs, one answer each, 360 in all: Perplexity (sonar), a ChatGPT-style engine (OpenAI gpt-5.4-mini with web search) and Claude (Haiku 5.5 with web search). These can answer differently from the consumer apps.
- Names in each answer were sorted by type (developer, portal, broker, media) by a model; 358 of 360 answers parsed. Developer names were matched by pattern, which over-counts a few broad names.
- 2,225 citations across 767 websites were labelled by a model from each domain and its home page title. The 90 most cited were checked by hand and 11 corrected. About 5% stayed “unsure”.
- The developer sample: 60 developers, 15 per city, each with one active residential project. Per city: 4 large listed (listed on the NSE or BSE), 6 mid-sized (unlisted, several projects or cities), 5 small regional (unlisted, one city, a handful of projects). Named and cited counts use only the 90 answers about the developer’s own city.
- One run, on one day. Answers change from day to day.
Project pages
- One live project page per sampled developer, checked on 26 October 2026, from the static page code, with a browser fetch where the page builds itself in the browser.
- Speed: Lighthouse 12.6.0, mobile, simulated slow 4G, headless Chrome on one computer. Lab data, not real-user data, and sensitive to the network. 57 pages measured, 3 errored.
- Each share’s base is the pages where the answer could be read. 11 pages build their form in the browser and count as unknown. 6 developers were replaced because their sites could not be read, tilting the sample towards sites that work without JavaScript. The chat widget count misses tools loaded through a tag manager.
Market, rules and benchmarks
- Market figures are ANAROCK’s, read in its own reports where marked; the 2025 city figures come from press coverage of its release. ANAROCK rounds to the nearest 50 units. Other research firms use other city sets and methods; we did not mix them.
- Rules are read from TRAI, MeitY and Meta’s own documents. For the 2026 amendment we read TRAI’s press release, not the full regulation, so clause numbers and start dates are not confirmed. State RERA advertising rules are left out: we could not reach the primary texts.
- The two speed-to-lead studies are from the US and are labelled as such. Neither tells you how Indian developers perform.
- No developer is named in this report. Portals and media sites are.
References
- Oldroyd, McElheran and Elkington, “The Short Life of Online Sales Leads”, Harvard Business Review, March 2011 US companies. Subscriber article; figures read from HBR’s own text.
- 2025 Online Homebuyer Mystery Shop Report, Blue Gypsy, Melinda Brody & Company and Denim Marketing (report page, published with Lasso CRM) US home builders.
- ANAROCK Research, “Indian Residential Real Estate: A Review and the Road Ahead”, January 2026
- ANAROCK year-end release on city sales, 26 December 2025, as reported by Business Today Secondary: press coverage of ANAROCK’s release.
- ANAROCK, Residential Market Viewpoints, Pan-India, Q3 2026
- PTI tally of listed developers’ FY26 sales bookings, via HDFC Sky, 24 June 2026 Secondary: a press tally of company disclosures.
- TRAI, Performance Indicators April to June 2026, Press Release No. 135/2026
- TechCrunch, WhatsApp confirms more than 400 million users in India, 26 July 2019 Secondary: a company statement reported by the press.
- TRAI, Telecom Commercial Communications Customer Preference Regulations, 2018
- TRAI, Telecom Commercial Communications Customer Preference (Second Amendment) Regulations, 2025
- TRAI Press Release No. 119/2026 on the Third Amendment Regulations, 2026, 18 September 2026
- The Digital Personal Data Protection Act, 2023
- Digital Personal Data Protection Rules, 2025, G.S.R. 846(E)
- WhatsApp Business Messaging Policy
- Google Lighthouse (version 12.6.0 was used for the speed tests)
- web.dev, Largest Contentful Paint: 2.5 seconds or less is “good”
Published 27 October 2026 by the Intend team at Kaigen Labs. To quote a figure, cite “Intend, India Real Estate Enquiry Report 2026” with a link to this page.