The Airtel giveaway multiplied Perplexity’s India user base by 5.4 times and its India revenue by 4.6 times. Divide one by the other and the company sells answers to Indian users for less per head today than it did before the promotion started.
Perplexity gave away a year of Perplexity Pro to Bharti Airtel’s 360 million customers, collected 56 million downloads in seven months, and peaked at 22 million monthly users in India. New app store revenue tracking from Sensor Tower and Appfigures shows India revenue climbing after the offer closed, which most outlets read as proof that the giveaway converted. Run the two datasets against each other and a different number falls out. Perplexity earned about 1.31 US cents per Indian monthly active user in January 2025, before the deal. In July 2026, after the largest consumer AI giveaway ever run in a single country, it earned about 1.11 cents. Airtel earns roughly $3.00 a month from the same handset.
What Happened
TechCrunch reporter Jagmeet Singh published the first cohort data on 18 August 2026, drawn from Sensor Tower and Appfigures estimates shared with the outlet.
The download numbers are enormous. Perplexity logged 5.9 million India downloads in July 2025, the launch month, up 625% from June and more in one month than the 5.4 million it had gathered across the whole first half of the year. Across the seven months Airtel customers could claim the offer, Sensor Tower counted 56 million India downloads, more than nine times the preceding seven-month stretch. Appfigures measured the same spike from a different angle: Perplexity averaged about 11,200 India downloads a day in the week before launch, about 223,000 a day in the first week after, and about 305,000 a day between mid-September and mid-October 2025.
Monthly active users doubled to 8.9 million in July 2025 and topped out near 22 million in October. Once Airtel closed new redemptions on 16 January 2026, downloads collapsed. Sensor Tower counted 3.3 million between February and July 2026, a drop of more than 90% from the six months before.
The users stuck better than the downloads. Perplexity held close to 14 million monthly actives in July 2026, down 37% from the October peak but more than five times the roughly 2.6 million it averaged in the first half of 2025. Sensor Tower analyst Abe Yousef told TechCrunch that usage stayed resilient after the promotion window closed.
Then the revenue line, which is what the industry seized on. Sensor Tower estimates India in-app purchase and subscription revenue between February and mid-August 2026 ran about 60% above the period when the Airtel offer was live. From roughly 18 July through 12 August, as the earliest free subscriptions hit their renewal dates, average daily India in-app revenue ran 9% above the previous 30 days and 27% above the first-half 2026 average.
Appfigures put dollar amounts on it. Perplexity’s monthly net mobile revenue in India went from about $34,000 in January 2025 to about $70,000 in December 2025 to about $156,000 in July 2026. The company took an estimated $878,000 in India across the first seven months of 2026, up 16% on all of 2025.
The Backstory
Airtel and Perplexity announced the partnership on 17 July 2025. All 360 million Airtel customers across mobile, broadband and DTH could claim a free 12-month Perplexity Pro subscription through the Airtel Thanks app. Airtel valued the giveaway at ₹17,000 a head, about $200. Pro buyers get frontier models from OpenAI, Anthropic and Google inside Perplexity, plus deep research, image generation and file analysis.
Aravind Srinivas got what he was buying. India runs on cheap data across more than a billion internet subscribers and over 700 million smartphones, and it had already become the largest single market for generative AI app downloads. It is also the market where those downloads produce the least money.
Airtel got something more specific. Indian telcos cannot keep raising tariffs, so they compete on bundled value, the same way they bundled Netflix, Amazon Prime and Disney+ Hotstar into prepaid packs across the previous five years. A $200 AI subscription costs the carrier close to nothing in cash and gives a subscriber a reason not to port out.
Reliance Jio and Google copied the structure within four months. Google agreed to hand 18 months of Google AI Pro, a ₹1,950-a-month plan valued at ₹35,100 over the term, to Jio’s 505 million telecom users, starting with 18 to 25 year olds. OpenAI moved a week earlier, making its ₹399-a-month ChatGPT Go plan free in India for a year from 4 November 2025. Sam Altman has said India is OpenAI’s second-largest market and may become its biggest.
Perplexity went first, so its cohorts hit the renewal wall first. Those subscriptions auto-renew a year from activation, which means Airtel customers who never opened a settings screen got billed.
The Plan
Perplexity’s plan is a freemium funnel stretched across a national carrier base. Give the premium tier away for twelve months, let people build the habit of asking Perplexity instead of Google, then charge when the free year ends. Srinivas has said as much about India being a shortcut in the race against OpenAI.
The strategy carries an unusual constraint at this company. Perplexity walked away from advertising in February 2026 and told the Financial Times it has no plans to bring it back, on the argument that sponsored placements poison the trust a citation-first answer engine sells. Ads had been running under 0.1% of revenue, so killing them cost almost nothing at the time. It also removed the only mechanism Perplexity had for earning money from a user who will not subscribe.
Compare that with the two rivals running the same India play. OpenAI switched ads on for ChatGPT Free and Go tiers in February 2026, the same month Perplexity switched them off. Google sells the largest advertising business on earth. Both can monetize an Indian user who takes the free year and then declines to pay. Perplexity has one price, $20 a month, and no fallback.
The Business Model Angle
Every write-up of this story ran the same arithmetic: 22 million peak users, a 60% revenue jump, therefore conversion works. One outlet speculated that 2% to 3% of the millions who tried Pro became paying subscribers, which would be tens of thousands of $200 subscriptions.
Test it against the money Appfigures actually measured.
Perplexity’s India mobile revenue in the first seven months of 2026 was $878,000, which annualizes to roughly $1.5 million. A 2% conversion on 22 million peak monthly users at $200 each would be $88 million a year. The 3% version would be $132 million. The observed number sits somewhere between 60 and 90 times below the floor of that guess.
Work backwards from the July 2026 figure instead. Appfigures reports net revenue, after the app stores take their roughly 30%, so gross India mobile billings ran near $223,000 that month. At the $20 Pro sticker, that supports about 11,100 subscriber-equivalents. Set that against 14 million monthly actives and you get 0.08%, or one payer for every 1,260 active users. Set it against the 56 million downloads collected during the offer window and it falls to one in roughly 5,000.
The renewal wave itself is smaller than the percentages suggest. July 2026 revenue averaged about $5,030 a day. A 9% lift over the prior 30 days is about $450 a day, which annualizes to roughly 1,000 subscriber-equivalents. Take the more generous framing, 27% above the first-half average, and the number reaches about 2,300. Somewhere between one and two thousand new annual subscriptions came out of a cohort that downloaded the app 5.9 million times in a single month.
Here is the figure nobody printed. Perplexity earned about $34,000 from India in January 2025 against roughly 2.6 million monthly actives, which works out to 1.31 cents per user. In July 2026 it earned about $156,000 against 14 million monthly actives, or 1.11 cents. Users grew 5.4 times. Revenue grew 4.6 times. Revenue per active user fell about 15%.
Now put Airtel’s side of the ledger next to it. In the quarter ended 30 June 2026, Airtel reported India mobile ARPU of ₹264, up from ₹250 a year earlier, alongside a record 1.0 million postpaid additions and 21.1 million more smartphone data customers year on year. At the rupee rate Reuters used through the India AI coverage, ₹264 is about $3.00 a month. Airtel monetizes the Indian handset roughly 270 times better than Perplexity does, using an asset Perplexity paid for in inference.
That is the trade both sides signed. Airtel handed over a marketing slot in the Thanks app and booked a retention lever at zero cash cost, and its ARPU rose through the giveaway period. Perplexity handed over twelve months of purchased frontier-model tokens resold at a price of zero, and came out the other side selling answers for a penny a head.
The subscription model works when the free period ends and price discovery begins. India’s price discovery already happened. OpenAI found ₹399 a month. Google found ₹1,950 with a telco paying the freight. Perplexity is asking $20, roughly ₹1,760, in a market where its own measured willingness to pay across 14 million users clears at about 13 cents a year.
The Risk
The case against this reading is real and worth stating.
App intelligence firms only see the app stores. Indian consumers pay for plenty of software through UPI on the web, and Perplexity sells directly through its own checkout, so the true India payer count sits above what Appfigures and Sensor Tower can measure. Enterprise and Comet browser revenue never touches those pipes either. Gross the mobile number up three or four times and the per-user figure improves, though the gap against Airtel’s $3.00 stays in the double digits.
The comparison also mixes a single month of revenue against a half-year average of users, because that is the shape of the data TechCrunch published. A cleaner cohort file would help. Ariel Michaeli at Appfigures made the honest point himself: the visibility from the promotion pulled in paying users who were never Airtel customers, so not all of the revenue growth belongs to the conversion story.
The auto-renew problem cuts the other way and cuts harder. Those free subscriptions renew unless the user cancels. Sensor Tower told TechCrunch it cannot separate a deliberate purchase from a missed cancellation, and it cannot separate former Airtel subscribers from anyone else. A revenue line built on people who forgot to cancel churns out in month two. The July renewal bump measures friction as much as it measures demand, and nobody will know which until the refund and cancellation data catches up.
One number does favour the bulls. Post-offer downloads still average about 18,000 a day, above the 11,200 Perplexity managed before Airtel ever picked up the phone. Habit formation left something behind, and 14 million monthly actives is a real asset if the company ever finds a way to charge them something other than $20.
Quick Questions
Did the Airtel deal fail? Not on its own terms. Perplexity wanted scale in India and got it, holding 14 million monthly users against 2.6 million before. The deal failed the test everyone is applying to it today, which is whether free premium AI converts to premium prices in a price-sensitive market.
Who paid for the giveaway? Perplexity. Airtel contributed distribution and a slot in its rewards app. Perplexity contributed twelve months of frontier-model inference it buys from OpenAI, Anthropic and Google and resold at zero.
How much money does Perplexity make in India? About $878,000 in mobile app revenue across the first seven months of 2026, per Appfigures, on a base of 14 million monthly active users. Web and enterprise revenue sit outside that figure.
What happens to the OpenAI and Google versions? The first ChatGPT Go cohort reaches its renewal date from 4 November 2026. The Jio cohort runs 18 months, which puts the first expiries around April 2027. Both companies can sell ads to users who decline to pay. Perplexity cannot.
Is $20 a month the wrong price for India? The measured evidence says it clears at a rate of about one paying user per 1,260 active ones. OpenAI priced ChatGPT Go at ₹399, roughly a fifth of Perplexity Pro. Perplexity has no India-specific tier.
The Business Model Analyst Take
Distribution deals do not create willingness to pay. They reveal it, and the reveal is usually unflattering.
Perplexity ran the cleanest natural experiment the consumer AI market has produced. One country, one carrier, 360 million eligible customers, twelve months of the full premium product at no charge, and a hard cutoff date that creates a proper cohort. The answer came back: 14 million Indians will use a frontier AI assistant every month, and roughly 11,000 of them will pay $20 for it.
The industry is reading the 60% revenue increase as validation because 60% of a small number still looks like growth on a slide. Founders should read the denominator instead. When you multiply users by 5.4 and revenue by 4.6, you have not built a business in that market, you have bought an audience and discovered its price. Perplexity discovered that its India price is about 13 cents a year per active user, against a carrier that pulls $36 a year out of the same person for moving their bits.
Anyone running the same playbook should copy Airtel, not Perplexity. Airtel monetized somebody else’s product launch, improved its ARPU and its postpaid mix through the promotion window, and never wrote a cheque. That is the position worth engineering: own the relationship and the billing rail, then let the venture-funded newcomer subsidize your retention.
For Perplexity the harder question is what the 14 million are for. With advertising off the table by choice, a user who will not pay $20 generates inference cost and nothing else. The company can build an India tier, restore some non-subscription monetization, or treat the cohort as a training and distribution asset it never intends to bill. Those are three different companies. The renewal dates rolling through January 2027 will force the choice, and the agentic browser push through Comet suggests Srinivas already knows the answer engine alone will not carry it.
Watch the November 2026 ChatGPT Go cohort. OpenAI priced its India tier at a fifth of Perplexity’s and kept an ad product live underneath it. If OpenAI converts at a materially better rate on the same population, the lesson was never about giving AI away in India. It was about charging Indian prices for it, which is the discipline Spotify learned across a decade of emerging-market ARPU and every AI company is now relearning at speed.
