State of AI 2026 Report form Sensor Tower: AI is becoming the new front door to shopping

AI’s consumer story used to be simple: people downloaded ChatGPT, asked weird questions, screenshotted the answers, ppl made fun of them on Reddit, and everyone moved on until a new model dropped and then we did it all again.

That phase is over.

Sensor Tower’s State of AI 2026 report shows a different market forming. AI assistants are becoming places where people discover products, compare options, see ads, click into retailers, and convert. The most important consumer AI battleground may be shopping, because shopping is where attention turns into money fast enough for everyone to notice.

ChatGPT still reached a giant milestone first. Sensor Tower says it became the fastest app ever to hit 1B monthly active users, reaching that mark in May 2026, three years after launch. That is faster than TikTok, YouTube, Instagram, Google Maps, Gmail, WhatsApp, and Facebook.

But the report’s more interesting point is that ChatGPT’s dominance is getting messier. Its True Audience market share fell below 50% for the first time in March 2026. By May, ChatGPT held 46% share across Sensor Tower’s 25-market analysis, ahead of Google Gemini at 28% and Claude at 10%.

The market is growing up. People are using more AI, paying for more AI, switching between AI tools more freely, and letting AI shape what they buy.

The AI app market is maturing, not fading

The lazy version of the story says AI app growth is slowing. The useful version says the market is shifting from installation to habit.

Sensor Tower projects generative AI app downloads will reach 2.3B in H1 2026, up 7% half-over-half. That is healthy growth, but much slower than the 26% growth in H2 2025 and 71% growth in H1 2025. IMO, that kinda tracks with early exposure. I think there will be another spike when the current era of "frontier AI" becomes more accessible to users at a lower cost, and that wave of people who tried once and gave up on AI realize again how useful and intelligent these tools have actually become.

And again, the money tells a different story. Global in-app purchase revenue from AI apps is projected to pass $4B in H1 2026, up 36% half-over-half.

Usage is even stronger. Sensor Tower projects AI app time spent will more than double year-over-year to 36B hours in H1 2026. Total sessions are projected to reach 965B, more than 100x H1 2023.

That means fewer easy download wins, but deeper usage from people who stayed.

The same pattern shows up on the web. Generative AI websites generated more than 67B visits and nearly 23B hours of usage globally in Q1 2026. Visits grew 28% year-over-year, while time spent grew 41%.

When time spent grows faster than visits, users are doing more than checking in. They are working, researching, shopping, comparing, and looping back.

ChatGPT is huge, but users are willing to switch

ChatGPT’s 1B-user milestone shows the company's dominant position. But it also creates a strange vulnerability: when a tool becomes default, every product decision gets political, commercial, and personal at massive scale.

Sensor Tower says ChatGPT’s US uninstalls spiked after OpenAI’s agreement with the Department of War, peaking roughly 200% above the app’s average during the week of March 9-15. Many users appeared to switch to Claude after Anthropic declined a Pentagon partnership. Claude recorded more daily downloads than ChatGPT from March 1-5.

ChatGPT later retook the daily lead, but the gap stayed narrower than before March.

That does not mean ChatGPT is in trouble. It means AI assistants are not as sticky as social graphs. Your friends are on Instagram. Your work history is in Gmail. Your shopping history is on Amazon.

Your chatbot can be replaced in five minutes if another one feels smarter, safer, cheaper, or less annoying that week. The apps of course are trying to reverse that, and for some users, it seems to be working.

Sensor Tower’s retention data points in the same direction. ChatGPT’s churned-user share rose from 12.7% in January to 14.5% in April, while Claude’s churn declined rapidly. Sensor Tower also notes that ChatGPT’s churn increase coincided with the introduction of ads, along with the Department of War controversy.

That is the consumer AI paradox: ChatGPT is the default, but the default is still negotiable.

Shopping is where AI starts looking like a platform

The most important section of the report is the shopping data.

Sensor Tower found that GenAI referrals to shopping websites increased across every major retail category between Q4 2024 and Q1 2026. Computers and consumer electronics led the trend, with the share of visits from GenAI sources rising nearly fourfold, from 0.2% to 0.8%.

That still sounds tiny. It is tiny if you only measure last-click traffic.

The report argues AI’s influence is bigger than direct referral share. AI assistants also shape brand visibility through citations, recommendations, and research answers before a shopper ever visits a retailer.

Computers and consumer electronics brands had a high share of AI citations, suggesting people use AI heavily for research before purchase. Home and Garden behaved differently: a higher share of GenAI referral traffic and a lower citation share suggest AI is driving more direct downstream visits there.

This is the new shopping funnel:

1. Ask an AI assistant what to buy.
2. Compare products inside the answer.
3. See which brands get cited.
4. Click to a retailer, or remember the brand later.
5. Convert after a longer research session.

Google Search trained brands to fight for blue links. AI assistants train them to fight for recommendation slots.

That is a much stranger game.

Amazon and Walmart are taking opposite paths

Retailers are already choosing sides.

Sensor Tower says Amazon has largely blocked ChatGPT web crawlers from scanning its product catalog. Target and Walmart, by contrast, have announced partnerships with ChatGPT.

The traffic data reflects that split. GenAI’s share of referral traffic to Walmart and Target climbed above 1.5%, while Amazon stayed around 0.5% in recent months. Temu and SHEIN lagged too, with roughly 0.3% of web traffic coming from AI sources.

Amazon’s strategy makes sense if you believe the customer relationship must stay inside Amazon. Amazon has its own assistant, Rufus, and the company does not need ChatGPT sitting between shoppers and product pages.

Walmart’s strategy makes sense if you believe AI assistants are becoming the new discovery layer. If the customer starts in ChatGPT, Walmart wants to be findable there.

Both strategies can work. They imply different futures.

In one version, every major retailer builds its own assistant and keeps the agent inside its walls. In the other, general-purpose AI assistants become shopping routers, sending traffic to whichever retailers make their catalogs legible, partner-friendly, and easy to transact with.

The first future favors Amazon. The second gives Walmart, Target, and category specialists a cleaner shot at intercepting demand.

Rufus users convert at nearly twice the rate

The strongest proof that shopping agents matter comes from Amazon’s Rufus data.

Sensor Tower says Amazon shoppers using Rufus converted at nearly twice the rate of non-Rufus shoppers. Rufus-assisted sessions stayed above 40% conversion throughout Q1 2026, compared with roughly 20% for non-Rufus shoppers.

Rufus-assisted shoppers also spent much longer before converting: about 40 minutes, versus 15 minutes for non-Rufus shoppers.

That sounds inefficient until you remember what shoppers are doing. They are comparing, checking details, asking follow-up questions, and working through complex purchase decisions.

For a cheap impulse purchase, AI may be overkill. For a laptop, TV, stroller, appliance, mattress, or holiday deal pile, an assistant can keep the shopper inside the retailer’s environment while answering the questions that would normally send them to Google, Reddit, YouTube, or a review site.

That is the real business case. AI shopping assistants are not only about answering questions. They are about keeping high-intent research from leaking out of the store.

Walmart’s Sparky is earlier, but growing faster. Sensor Tower says the share of Walmart monthly active users using Sparky rose 27% between December 2025 and April 2026. Average daily active users rose nearly 50%. Walmart also reports weekly active users are up over 100% in the last quarter, and shoppers using Sparky have average order values about 35% higher than non-users.

Retailers do not need every shopper to use an assistant. They need the valuable shoppers to use it when the cart is large, confusing, or comparison-heavy.

The creator side is shifting too

Sensor Tower’s report shows AI assistants becoming a new front door for discovery: shoppers ask AI for product help, retailers build agents into their apps, and advertisers start testing ChatGPT as a research-stage ad channel.

Adobe’s latest creator data shows the other half of that shift: the supply of content is also being rebuilt around AI.

In Adobe’s 2026 Creators’ Toolkit Report, 87% of creators who use creative AI said it is growing their business or audience, and 75% described it as integrated or essential to their workflow. That tracks with our recent Neuron piece on the report: AI is no longer a side experiment for creators. It is becoming creative infrastructure.

But the tension is the same one Sensor Tower found in shopping and ads: more AI does not automatically mean more trust or visibility.

Adobe found that among creators who say it is harder to stand out today than a year ago, 53% blame sheer content volume, while 42% say AI-generated content makes it harder for unique voices to break through. At the same time, 58% said their ability to compete with larger teams or studios feels stronger since using creative AI.

That is the new internet squeeze. AI helps creators produce more, while AI assistants change how audiences find what to watch, buy, and trust. The winners are unlikely to be the people who generate the most stuff. They will be the creators and brands with clearer voice, better provenance, stronger human judgment, and AI systems that keep control close to the person doing the work.

Adobe’s most useful stat may be the control stat: 85% of creators said the final creative decision should always remain with the creator, and 90% said it is important to obtain copyright protection for work made with AI assistance.

Put that next to Sensor Tower’s data on Rufus, Sparky, ChatGPT ads, and AI shopping referrals, and the pattern gets sharper: AI is becoming the operating layer between creation, discovery, and conversion. But the scarce resource is not output anymore. It is trust.

ChatGPT ads are the next logical step

Once AI becomes a place for discovery, ads arrive.

Sensor Tower says ChatGPT began testing ads with select US users in February 2026, then expanded availability in March and April. By the end of May, the number of ads and the number of users seeing ads had each increased roughly 7x compared with the first week of March.

The daily share of users seeing ads remained relatively low. The direction is still clear. Sensor Tower frames this as the early stage of a broader ad system, with OpenAI introducing tools such as conversion optimization and campaign management features.

The first advertisers tell you what ChatGPT is becoming.

Shopping and Software brands accounted for nearly half of all ads on ChatGPT in Sensor Tower’s US Android sample. Shopping alone made up 33.3% of ChatGPT ad impressions, while Software made up 15.8%. Leading subcategories included Home and Garden, Apparel, general Shopping, food delivery, consumer electronics, news, education, sports and outdoors, and travel booking.

That mix is different from the social platforms. LinkedIn is software-heavy. Pinterest is shopping-heavy. TikTok and Instagram lean harder into consumer packaged goods and media.

ChatGPT’s ad mix looks like research and consideration.

That is where the money is. A user asking “what laptop should I buy for video editing under $1,500?” is more valuable than a user doomscrolling past a laptop ad while half-watching TV.

AI ads will feel different because the ad slot may sit inside the decision process, not beside it.

That is powerful. It is also risky.

The risk is trust

The counter-case is simple: ads can poison the thing that made AI assistants useful.

People ask AI assistants for help because the interface feels more like advice than media. Search ads already taught users to squint at sponsored results. Social ads taught users to scroll past persuasion. AI ads could teach users to distrust answers.

That risk gets sharper in shopping. If an AI assistant recommends a product, users need to know whether the recommendation came from relevance, sponsorship, commission, inventory, margin, or some messy blend of all five.

OpenAI can build ad labels. Retailers can build guardrails. Regulators can set disclosure rules. The user’s trust will be harder to restore if the first version feels slippery.

There is also a measurement trap. Sensor Tower notes that GenAI sources still account for less than 1% of total web traffic in shopping. The influence may be meaningful, but the hard attribution is early. Citations, research behavior, and later purchases are more difficult to value than clean last-click traffic.

Brands will still spend into it because marketers follow attention. The hard part will be proving which AI mentions actually moved revenue.

AI is becoming the new shelf

Sensor Tower’s broader advertising data shows the same migration.

US digital ad spend on creatives featuring AI-related terms reached $1.3B, up 48% year-over-year. Impressions climbed to 167B, up 46%. Health and Wellness led major categories with a 165% increase in ad spend on AI-related creatives. Financial Services, Business and Industrials, and Media and Entertainment each grew more than 60%.

Best Buy was the top retailer for retail media ads mentioning AI between May 2025 and April 2026, generating more than 2B impressions. The product list included Meta Ray-Ban glasses, Microsoft Surface, QuickBooks, Dell PCs, Samsung Galaxy, Google Pixel, HP, and Lenovo.

AI is becoming a product attribute, a search keyword, an ad claim, and a recommendation surface at the same time.

Sensor Tower also found that US app downloads driven by searches containing “AI” grew 651% in Health and Wellness and 536% in Financial Services between Q1 2024 and Q1 2026. Business and Productivity Software and Jobs and Education more than doubled.

That means “AI” is no longer only a category. It is a discovery behavior.

People are not only downloading AI apps. They are searching for AI fitness, AI finance, AI education, AI productivity, AI design, AI shopping, AI dating, AI everything. Some of those searches will lead to useful products. Some will lead to feature confetti.

Brands will overuse the label because it works. Consumers will get more selective because they have to.

What to watch next

Three signals matter from here.

1. First, watch whether AI shopping traffic breaks out of the sub-1% zone. If referrals from ChatGPT, Gemini, Perplexity, and other assistants keep climbing, retailers will start treating AI visibility like SEO.
2. Second, watch whether retailers open or close their catalogs. Amazon can afford a walled-garden AI strategy. Smaller retailers probably cannot. The more product data an assistant can see, the more likely it can recommend, compare, and convert.
3. Third, watch whether ads change retention. Sensor Tower already links ChatGPT’s modest churn increase to the period when ads were introduced, alongside the Department of War controversy. That does not prove ads caused churn. It does prove users are willing to move when trust or experience changes.

The open question is whether AI assistants can monetize discovery without making the advice feel bought.

Search managed that bargain for two decades, imperfectly but profitably. Social platforms managed a different bargain, trading attention for targeting. AI assistants are trying to monetize something more delicate: the moment when a user asks for help.

If they get it right, AI becomes the new front door to commerce.

If they get it wrong, the most valuable users may do what Sensor Tower says they already do: switch assistants.