Hostinger survey: Established side businesses risk getting left behind by newcomers using AI in the US

In America’s side-business economy, the most seasoned operators are the least likely to adopt AI. The pattern holds even when you control for income and business type – except in one revealing category, our survey shows.

Economists studying the use of artificial intelligence across different industries and geographies have identified a persistent puzzle. In controlled experiments – where workers are given AI tools, trained to use them, and assigned specific tasks – the technology is equalizing. Less experienced workers benefit the most, and the gap between top and bottom performers narrows.

But step outside the laboratory and a different picture emerges. What happens in an economy where there is no IT department, no learning-and-development program, no manager nudging workers toward new tools?

The results are counterintuitive and are a natural test case for AI adoption without institutional support.

Hostinger’s survey data, based on 4,002 digitally active workers in the United States and India, found that Americans who have operated a side job for 3 years or more – before the public emergence of ChatGPT – adopt AI tools at a substantially lower rate than new entrants into their market, with significant impact on earnings.

The survey defines a side business as any income-generating activity outside a primary job – freelancing, online selling, content creation, tutoring, handyman services – and asks whether operators use AI tools to support it.

The US side-business economy is large, with estimates ranging from a record 9.3 million formal multiple jobholders to 72.9 million independent workers, and 27% of American adults reporting a side business. This market exhibits a wide income gap, where the average monthly side-job income of $885 dramatically exceeds the median of just $200, a split between high and low earners that maps closely onto our report’s central finding about who adopts AI and who does not.

Among US side-business owners who have operated for more than three years, just 46% use AI tools – a drop of 16 to 21 percentage points below every other tenure cohort, which clusters between 63% and 67%. The decline is not gradual. It is a cliff, concentrated entirely among the most experienced operators.

Two obvious explanations come to mind. Maybe established operators run businesses where AI just isn’t that useful – handyman work, skilled trades, crafts. Or maybe they can’t justify the cost of a subscription. The data rules out both.

If you’ve been freelancing for three years, your newer competitors are using AI at rates 15 points higher than you. That’s the same work, the same clients, the same market.

This table eliminates the business-mix explanation. A freelancer who has been operating for three years and a freelancer who started six months ago are doing the same type of work, serving the same type of client, competing in the same market. Yet the established freelancer adopts AI at 62% while the newer one does so at 76–79%.

That 15-point gap is not about the nature of freelancing. It is about the operator. The same pattern holds among online sellers (13-point gap), real estate operators (14 points), handyman services (12 points), and crafts (13 points).

Something about having run a business before AI existed makes people less likely to adopt it now – even when their direct competitors are doing so.

The income data confirms it. Among side-business owners earning $1,000 or more per month – people with both the revenue and the demonstrated capacity to run a profitable venture: 3+ year operators still adopt AI at just 60%, versus 70–84% for newer cohorts. Among those earning $500–999, the gap widens: 40% for established operators against 57–74% for the rest. These are not people who lack the resources for a $20/month ChatGPT subscription. They have chosen not to use one.

The one category that breaks the pattern

Among content creators – YouTubers, bloggers, TikTok producers, newsletter writers – established operators adopt AI at 84%, the highest rate of any tenure-business combination in the dataset. This is the one sector where generative AI doesn’t just augment the work – it threatens to replace it. A freelance writer who has been producing blog posts for four years watched ChatGPT learn to do the same thing in 2023. A handyman who has been fixing plumbing for four years has not had the same experience. When AI visibly threatens the core work, experience doesn’t slow adoption. It accelerates it.

The mechanism becomes clearer in the barrier data. Among established non-adopters – the people running side businesses for more than three years who do not use AI – the dominant obstacle is not cost (1%), not lack of training (6%), not uncertainty about how to start (6%). It is the belief that AI simply does not apply to them.

Fully 52% of established non-adopters say their business “doesn’t benefit from AI.” Among non-adopters who started in the past year, the figure ranges from 28% to 31%. These operators have not been priced out of AI. They have not failed to learn it. They have decided it is irrelevant. The income data – where AI adoption correlates with earnings at every level measured – suggests this belief is wrong. And it becomes a cycle: the longer someone operates without AI, the more their workflows optimize around its absence, and the more reasonable the conclusion feels from the inside.

India adopted AI everywhere

In comparison, among Indian side-business owners, AI adoption runs between 87% and 93% across every tenure cohort. Operators who have been running ventures for more than three years adopt at 92% – virtually identical to those who started last month. The 20-point American experience penalty simply does not exist.

India runs similar types of side businesses. If business type were the reason, you’d expect to see the same gap there. You don’t. One likely reason: India’s digital business infrastructure is newer. Fewer Indian entrepreneurs built workflows in the pre-AI era, so there are fewer legacy processes to unwind. American incumbents, by contrast, spent years optimizing around tools that worked well enough – and now face the real cost of abandoning systems that are still functional, if no longer fully competitive.

Imas and Shukla lay out two plausible futures for AI adoption. In one – “catch-up” – adoption broadens through falling prices, better tools, and wider access, and AI’s equalizing potential eventually shows up in the aggregate data. In the other – “lock-in” – a widening skills gap gets harder to close.

India, where adoption is near-universal regardless of experience, looks like catch-up.

The United States, where a 20-point tenure gap persists even among high earners in the same business categories, looks like the early stages of lock-in.

The income data for established operators sharpens the picture. Three-plus-year operators have the highest share of $2,000+/month earners of any cohort: 18.7%. But they also have the most people stuck in the $100–499 range: 32%. The cohort is split in two. A small minority is thriving – possibly the 46% who did adopt AI. The majority are not. And their overall share of $1,000+/month earners (37%) has been overtaken by operators who started just one to three years ago (48%).

For workers in corporate settings, this finding should register as a warning rather than a curiosity. The side-business economy is what happens when AI adoption proceeds with no institutional support – no training program, no mandate from leadership, no protected time to learn. The result is not broad adoption or gradual convergence. It is a sharp split along the axis of experience – precisely the quality that professionals most rely on to justify their market position.

The content creators in this data didn’t wait for a perfect moment. They saw AI moving into their space and moved with it – ahead of the curve, not behind it.

The good news for everyone else: the adoption window is still open. Unlike content creators, most side-business owners haven’t faced direct AI competition yet – which means there’s still time to build the habit before it becomes urgent.

The operators pulling ahead right now aren’t necessarily more tech-savvy. They’re just the ones who tried something new before they had to.

Methodology

Survey design. Hostinger surveyed 4,002 individuals – 2,000 in the United States and 2,002 in India – via Cint between January 20 and 28, 2026. Respondents were recruited from online research panels with quota controls for age, gender, and geographic distribution within each country. The sample targets the digitally active working population – individuals who join online panels are by definition more likely to pursue digital side-business opportunities.

The margin of error is ±2.2% at 95% confidence for the full 4,002-person sample. Margins are wider for subsamples: approximately ±3.1% for country-level analyses (n≈2,000), and approximately ±3.1% for the US side-business-owner subset (n=1,014). All cross-tabulations in this report are based on the US respondent-level dataset (n=1,014 side-business owners).

Replication. The underlying respondent-level data is available upon request. Researchers and journalists seeking to verify or extend the analysis should contact Hostinger’s communications team at press@hostinger.com.