India virtually batted Sri Lanka out of the second Test by crossing the 600-run mark for the second game in a row, setting themselves up for another big win on a pitch that is offering plenty to the spinners.
India, resuming the day at 344 for three, went on to declare at 622 for nine post tea, thanks to half centuries from Ravichandran Ashwin (54 off 92), Wriddhiman Saha (67 off 134) and Ravindra Jadeja (70 not out off 85).
Overnight centurions Cheteshwar Pujara (133) and Ajinkya Rahane (132) could not add much to their overnight score.
In reply to India's mammoth total, Sri Lanka were left reeling at 50 for two at stumps with Ravichandran Ashwin taking both the wickets.
The hosts face a daunting task of saving the game and also the series with Ashwin and Jadeja making the ball talk on a helping surface.
"The Bounce is quite variable. Some balls are bouncing and few balls are keeping low. We knew that if they play sweep shot it's good for us we have the chance to get wickets," said Rahane after close of play.
"But it s not easy for batsmen playing on this wicket.
Tomorrow (5) will be key to our bowlers to bowl in right areas for long period of time."
He further said the team is not thinking of enforcing the follow on yet.
"We are not thinking as of now because as I said, it is not going to be easy for our bowlers as well. They have to bowl in right areas for longer period of time.
"Yes, the wicket is turning and it is not easy for batsmen. But for the bowlers, we have to bowl well and get their eight wickets. Let s see, we will wait and see tomorrow," Rahane added.
Ashwin, especially, looked well and truly back in his elements and bowled brilliantly to the left-handers.
He dismissed opener Upul Tharanaga for a duck as the southpaw whipped one straight into the abdomen of Rahul, who did well to latch on to the ball.
A little later, Ashwin sent back Tharanga's opener partner Dimuth Karunaratne (25) with the one that took the outside edge on way to the first slip.
Skipper Dinesh Chandimal (8) came to the crease and looked aggressive while batting alongside Kusal Mendis (16) as Sri Lanka avoided further loss until close of play.
Earlier, the middle and lower order fired to take India beyond 600.
Ashwin, batting at number six, struck his 11 Test fifty on his way to becoming the fourth Indian to complete a double of 200 wickets and 2000 runs.
He also became the fourth quickest in Test history to get to the milestone in 51 Tests, after Ian Botham (England, 42 Tests), Kapil Dev (India, 50 Tests) and Imran Khan (Pakistan, 50 Tests).
Post tea, Jadeja and Saha looked to score some quick runs. The left-hander smashed three fours and a six as he raced to his eighth Test half-century off 70 balls.
At the other end, Saha looked to attack as well, but he was stumped off Rangana Herath (4-154) in the 153rd over. It was the first instance of six half-centuries in one innings in a Test on Lankan soil and the second time for India after 2007 at the Oval.
It brought Mohammed Shami (19) to the crease who also threw his bat around. Jadeja-Shami put on 30 quick runs for the ninth wicket. Shami was last man out, caught at deep mid- wicket off Herath again. Umesh Yadav remained unbeaten on 8 as India crossed the 600-mark in the 156th over.
Earlier, Ashwin and Saha scored half-centuries as India reached 553/7 at tea.
But he couldn't build on it, as Herath bowled him in the 122nd over. Thereafter, Saha and Hardik Pandya (20) put on 45 runs for the seventh wicket.
Both batsmen survived DRS reviews for lbw, and took India past the 500-mark in the 134th over. The visitors thus became the first team to score 500-plus runs in successive Tests on Lankan soil.
Pandya was caught at long -off off Malinda Pushpakumara (2-156). But Saha carried on and put on a resolute 57-run unbeaten partnership with Jadeja.
In doing so, he scored his 5th Test half-century off 113 balls and brought up India s 550 just before the tea break.
This was after India reached 442/5 at lunch as both Pujara and Rahane fell in the morning session.
Sri Lanka were dealt some bad news in the morning when pacer Nuwan Pradeep (0-63) was ruled out of the remainder of the Test.
The pacer had injured his hamstring on day one and consequently the hosts have been left without a proper fast bowler in their attack for this game.
Even so, Pujara (overnight 128) couldn't take advantage of this situation as Karunaratne (1-31) trapped him lbw in the second over of the day. The bowler needed DRS review to get the dismissal in his favour.
Overall, Pujara faced 232 deliveries and hit 11 fours as well as a six. It brought an end to his 217-run partnership for the fourth wicket with Rahane (overnight 103).
The latter though continued batting resolutely and helped India across 400 in the 106th over. In doing so, he also brought up the 50-partnership with Ashwin for the fifth wicket off 84 balls.
Rahane looked set for a bigger score, but then against the run of play, he was out stumped off Pushpakumara in the 111th over. It was the first Test wicket for the debutant in his 100th First Class match. Overall, Rahane faced 222 deliveries and hit 14 fours.
Saha then joined Ashwin in the middle, and the duo put on 29 runs for the sixth wicket without any trouble, as India progressed towards another tall first innings score.
AI can make thousands of podcast episodes every week with very few people.
Making an AI podcast episode costs almost nothing and can make money fast.
Small podcasters cannot get noticed. It is hard for them to earn.
Advertisements go to AI shows. Human shows get ignored.
Listeners do not mind AI. Some like it.
A company can now publish thousands of podcasts a week with almost no people. That fact alone should wake up anyone who makes money from talking into a mic.
The company now turns out roughly 3,000 episodes a week with a team of eight. Each episode costs about £0.75 (₹88.64) to make. With as few as 20 listens, an episode can cover its cost. That single line explains why the rest of this story is happening.
When AI takes over podcasts human creators are struggling to keep up iStock
The math that changes the game
Podcasting used to be slow and hands-on. Hosts booked guests, edited interviews, and hunted sponsors. Now, the fixed costs, including writing, voice, and editing, can be automated. Once that system is running, adding another episode barely costs anything; it is just another file pushed through the same machine.
To see how that changes the landscape, look at the scale we are talking about. By September 2025, there were already well over 4.52 million podcasts worldwide. In just three months, close to half a million new shows joined the pile. It has become a crowded marketplace worth roughly £32 billion (₹3.74 trillion), most of it fuelled by advertising money.
That combination of a huge market plus near-zero marginal costs creates a simple incentive: flood the directories with niche shows. Even tiny audiences become profitable.
What mass production looks like
These AI shows are not replacements for every human program. They are different products. Producers use generative models to write scripts, synthesise voice tracks, add music, and publish automatically. Topics are hyper-niche: pollen counts in a mid-sized city, daily stock micro-summaries, or a five-minute briefing on a single plant species. The episodes are short, frequent, and tailored to narrow advertiser categories.
That model works because advertisers can target tiny audiences. If an antihistamine maker can reach fifty people looking up pollen data in one town, that can still be worth paying for. Multiply that by thousands of micro-topics, and the revenue math stacks up.
How mass-produced AI podcasts are drowning out real human voicesiStock
Where human creators lose
Podcasting has always been fragile for independent creators. Most shows never break even. Discoverability is hard. Promotion costs money. Now, add AI fleets pushing volume, and the problem worsens.
Platforms surface content through algorithms. If those algorithms reward frequency, freshness, or sheer inventory, AI producers gain an advantage. Human shows that take weeks to produce with high-quality narrative, interviews, or even investigative pieces get buried.
Advertisers chasing cheap reach will be tempted by mass AI networks. That will push down the effective CPMs (cost per thousand listens) for many categories. Small hosts who relied on a few branded reads or listener donations will see the pool shrink.
What listeners get and what they lose
Not every listener cares if a host is synthetic. Some care only about the utility: a quick sports update, a commute briefing, or a how-to snippet. For those use cases, AI can be fine, or even better, because it is faster, cheaper, and always on.
But the thing is, a lot of podcast value comes from human quirks. The long-form interview, the offbeat joke, the voice that makes you feel known—those are hard to fake. Studies and industry voices already show 52% of consumers feel less engaged with content. The result is a split audience: one side tolerates or prefers automated, functional audio; the other side pays to keep human voices alive.
When cheap AI shows flood the market small creators lose their edgeiStock
Legal and ethical damage control
Mass AI podcasting raises immediate legal and ethical questions.
Copyright — Models trained on protected audio and text can reproduce or riff on copyrighted works.
Impersonation — Synthetic voices can mirror public figures, which risks deception.
Misinformation — Automated scripts without fact-checking can spread errors at scale.
Transparency — Few platforms force disclosure that an episode is AI-generated.
If regulators force tighter rules, the tiny profit margin on each episode could disappear. That would make the mass-production model unprofitable overnight. Alternatively, platforms could impose labelling and remove low-quality feeds. Either outcome would reshape the calculus.
How the industry can respond through practical moves
The ecosystem will not collapse overnight.
Label AI episodes clearly.
Use discovery algorithms that reward engagement, not volume.
Create paywalls, memberships, or time-listened metrics.
Use AI tools to help humans, not replace them.
Industry standards on IP and voice consent are needed to reduce legal exposure. Platforms and advertisers hold most of the cards here. They can choose to favour volume or to protect quality. Their choice will decide many creators’ fates.
Three short scenarios, then the point
Flooded and cheap — Platforms favour volume. Ads chase cheap reach. Many independent shows vanish, and audio becomes a sea of similar, useful, but forgettable feeds.
Regulated and curated — Disclosure rules and smarter discovery reward listener engagement. Human shows survive, and AI fills utility roles.
Hybrid balance — Creators use AI tools to speed up workflows while keeping control over voice and facts. New business models emerge that pay for depth.
All three are plausible. The industry will move towards the one that matches where platforms and advertisers put their money.
Can human podcasters survive the flood of robot-made showsiStock
New rules, old craft
Machines can mass-produce audio faster and cheaper than people. That does not make them better storytellers. It makes them efficient at delivering information. If you are a creator, your defence is simple: make content machines cannot copy easily. Tell stories that require curiosity, risk, restraint, and relationships. Build listeners who will pay for that difference.
If you are a platform or advertiser, your choice is also simple: do you reward noise or signal? Reward signal, and you keep what made podcasting special. Reward noise, and you get scale and a thinner, cheaper industry in return. Either way, the next few years will decide whether podcasting stays a human medium with tools or becomes a tool-driven medium with a few human highlights. The soundscape is changing. If human creators want to survive, they need to focus on the one thing machines do not buy: trust.
By clicking the 'Subscribe’, you agree to receive our newsletter, marketing communications and industry
partners/sponsors sharing promotional product information via email and print communication from Garavi Gujarat
Publications Ltd and subsidiaries. You have the right to withdraw your consent at any time by clicking the
unsubscribe link in our emails. We will use your email address to personalize our communications and send you
relevant offers. Your data will be stored up to 30 days after unsubscribing.
Contact us at data@amg.biz to see how we manage and store your data.