Your feed is full of 1980s photos. It is also an onboarding lesson.

The interesting number in the AI image trends is not how many images were made. It is how many of the people making them had never used the app before.

When Google’s Nano Banana model went viral last September, Josh Woodward, VP of Google Gemini and Google Labs, said the app had picked up 23 million first-time users in about two weeks. Not 23 million sessions. First-time users. During the Ghibli wave in March 2025, Sam Altman said ChatGPT added one million users in a single hour, against five days for the first million at launch.

These were people who had heard of AI, had never touched it, had no plan to, and installed an app that week anyway. That is the hardest thing any business ever does, and it happened at a scale no advertising budget produces.

So the question worth asking is not what the trend was. It is what made a stranger try something for the first time, and whether a shop in Trivandrum can build the same conditions.

I think it can. There are five of them and none require technology.

One. The recommendation arrived sideways.

Nobody installed the app because the company told them to. They saw their cousin’s photo. Their colleague’s photo. Fifteen photos in a row from people they already trusted.

The business was not in the conversation. That is the point. A recommendation from a person carries weight that no amount of your own advertising can buy, because you are not the one making the claim.

Most local businesses spend their entire marketing budget speaking in their own voice. Almost none of them spend anything on making it easier for a customer to speak in theirs.

Two. They saw the result before they started.

This is the condition most businesses miss.

Nobody who uploaded a photo was wondering what would happen. They had already seen forty examples. The uncertainty was zero. They knew exactly what they were getting, which meant the decision to try cost them nothing mentally.

Compare that with what a local business usually offers a new customer. Come in for a consultation. Call us for a quote. Visit our showroom. In every one of those, the person has no idea what they are walking into, how long it will take, or whether they will be pressured. Uncertainty is a cost, and most first-try offers are full of it.

Three. The first action was one step.

Upload a photo. Paste a prompt. That is the whole thing.

It was not “learn AI.” It was not “set up an account and explore the features.” It was one action with one outcome, and it took under a minute.

If your first try has three steps, most people stop at the second.

Four. It cost nothing and committed them to nothing.

Free, instant, and no obligation to ever come back. Nobody had to speak to a salesperson. Nobody had to give a phone number before seeing the result.

Ask what your equivalent is, and be honest about whether it really is free of commitment. A free consultation where someone then has to say no to a person’s face is not free. It costs awkwardness, and people avoid it.

Five. Finishing it produced something they showed someone.

This is the part that closes the loop.

The output of the first try was an image. People posted it. That post became condition one for the next person. The trend did not spread because the companies pushed it. It spread because completing the first try automatically recruited the next stranger.

That is not a funnel. It is a loop. And it only works because the thing produced at the end was worth showing.

A good first try is not a funnel. It recruits the next person by itself.

What this looks like on a Trivandrum street.

The five conditions are the test. Run your current first-time offer against them and most fail on three, four or five.

A two-wheeler workshop. Bad first try: bring the bike in and we will look at it. Good first try: a ten minute free check, and the mechanic sends photographs of what he found on WhatsApp with a plain explanation in Malayalam. One step, no commitment, and the result is forwardable. People send those photos to their brother, their husband, the friend who knows bikes. Condition five happens without asking anyone to post anything.

A tailoring shop or boutique. Bad first try: bring the material and place an order. Good first try: walk in, get measured properly, take the measurement card home, order nothing. Ten minutes. Anyone who has had a churidar or a blouse stitched knows that measurements taken correctly are the whole difference, and that redoing them somewhere else is a real cost. The shop now holds something valuable and asked for nothing in return.

An entrance coaching centre. Bad first try: attend a demo class. Good first try: a forty minute diagnostic test, free, walk-in, with a printed sheet showing exactly which chapters the student is weak in. The parent leaves holding a document. In this city that document gets forwarded before evening. That is condition five doing all the work by itself.

A bakery or tea shop. Bad first try: a discount on a large order. Good first try: one item, small, cheap, and made well enough that people photograph it before they eat it.

An AC or appliance service. Bad first try: call us for a quote. Good first try: send a photo of the unit on WhatsApp and get back a plain answer and a price range, with no visit booked and nobody calling you afterwards.

Notice what all of these have in common. The customer leaves holding something. Not a promise, not a brochure. A photo, a card, a score sheet, a plate. That object is what makes condition five possible at all.

The objection.

The fair pushback is that a free first try just attracts people who will never pay.

Some of it does. That is a real cost and I would not pretend otherwise. But the trends showed something worth weighing against it. Both companies kept a large share of the people who arrived for a novelty. OpenAI reported 800 million weekly active users by October 2025, up from 500 million at the end of March.

They kept them because there was a second reason to come back that had nothing to do with the trend. That is the condition I would check before running any of this. If a customer’s first try is good but there is no obvious second thing, you have built a giveaway, not an onboarding.

Numbers I would not use.

Anything quantifying the trend running this month. The 1980s portrait trend is days old. No app-intelligence firm has published a measurement. Every engagement figure you see about it this week is invented.

“23 million first-time users.” This is Google reporting on Google, said at a Google media session. It is the most on-point figure in this post and it is still a vendor claim, not an audited one.

“One million users in one hour.” A founder’s post on X during his own growth moment. Read it as scale, not as a metric.

“Gemini downloads rose 45%, with India leading at 60%.” The 45% is real and traceable to Appfigures. The India split of 60% has no source I could find anywhere.

Almost all of this is global or US app-store data. The India-specific claims come from Google employees, so they are not independent.

Who this does not apply to.

If what you sell is genuinely high-stakes and slow, a free first try can work against you. Nobody wants a discounted trial of surgery or a sample of legal representation. In those cases the first try is not the product. It is a piece of the thinking: one clear answer to one real question, given away properly.

The five conditions still hold. Only the object changes.

What I would do first.

One. Write down what a new customer has to do to try you today. Every step, honestly.

Two. Run it against the five conditions. Most first-try offers fail on step count and on commitment.

Three. Cut it to one step, and remove anything that requires the customer to say no to a person.

Four. Make sure they leave holding something. That object is the only reason condition five ever happens.

None of this is a campaign. It is a redesign of the first ten minutes, and it is usually free.

The AI companies did not invent any of this. They stumbled into all five conditions at once and then, once they understood what had happened, built it into the product on purpose. A local business can do it deliberately from the start, which is an advantage, not a disadvantage.

At NexGU we build software for local businesses and MSMEs, and a lot of the work is exactly this. Not campaigns. Redesigning the first ten minutes so a stranger can get to a result without talking to anyone. If that sounds like your problem, write to me at gokul@nexgu.com or reach us at nexgu.com/connect.

Sources.

• Josh Woodward, VP Google Gemini and Google Labs, 8 September 2025, via Notebookcheck. 23 million first-time users and over 500 million images since the Nano Banana model launched. Company-reported, not independently audited. notebookcheck.net
• Cybernews and Axios, 31 March 2025. Sam Altman on adding one million users in one hour during the Ghibli trend. Founder statement on X. cybernews.com, axios.com
• Appfigures, via TechCrunch, September 2025. Gemini app downloads: 12.6 million in the first half of September against 8.7 million for all of August, a 45% month-over-month rise. Source for the chart. techcrunch.com
• Jagmeet Singh, TechCrunch, 17 September 2025. David Sharon of Google DeepMind on India leading Nano Banana usage and the retro trend being local to India. Google speaking about Google. techcrunch.com
• Inside AI, 9 September 2026. Reporting on the current 1980s ChatGPT portrait trend and its Bollywood variant in South Asia. insideai.news
• TechCrunch, October 2025. Altman’s 800 million weekly active users, against 500 million at the end of March. Source for the retention argument. techcrunch.com

The research above is not ours. It belongs to Appfigures, the reporters who published it, and Google and OpenAI for their own figures. Every one is linked.