About half of AI users have abandoned a purchase or switched brand because of an AI search answer | an Itera survey

Itera, Inc. (head office: Shinjuku-ku, Tokyo; Representative Director: Takayuki Muto) ran two surveys: one on shopping, among 2,000 men and women aged 18 to 70 across Japan, and one on the use of AI in shopping, among 300 people with experience of using generative AI for e-commerce and mail order.

33.9% have used generative AI when shopping. Among those users, 45.7% were shown a product they did not know about, and 33.6% of those went on to buy it.

Key findings

• 34.0% used generative AI or AI search when choosing a product or service in the past year (n=2,000): “many times” 19.2%, “once or twice” 14.8%
• The most researched category with generative AI was consumer electronics and gadgets at 39.6%, followed by travel and hotels at 31.5% and restaurants at 28.0% (n=679, multiple answers)
• Among those using AI for e-commerce, 45.7% had been shown a product, brand or shop they did not know about (n=300)
• Of those, 76.6% looked into it further and 33.6% actually bought it (n=137, multiple answers)
• 50.3% in total have abandoned a purchase or switched to another brand because of an AI answer (n=300)
• The triggers were negative statements about reputation (42.4%), quoted negative reviews (39.1%) and an unfavourable assessment in comparison with other products (33.1%) (n=151, multiple answers)
• 17.0% used the AI answer as their basis for deciding as it stood; 83.0% checked another source (n=300)
• Where an AI answer conflicted with the official site, 42.0% trusted the official site, 26.3% reviews on the e-commerce site and 8.3% the generative AI answer (n=300)
• 72.0% think companies and shops should understand and improve how generative AI answers about them (n=300)

The findings in detail

1. 33.95% have used generative AI or AI search when shopping (n=2,000)

We asked 2,000 men and women aged 18 to 70 across Japan whether, in the past year, they had gathered information using generative AI or AI search such as ChatGPT or Gemini when choosing a product or service.

Combining “many times” and “once or twice” gives 679 people, or 33.95%. In our August 2026 surveys, 26.0% had asked generative AI during a job search (n=880) and 23.5% when looking for a home (n=507) — shopping exceeds both.

2. The most researched category was consumer electronics and gadgets at 39.6% (n=679, multiple answers)

Consumer electronics and gadgets came first at 39.6%, travel and hotels second at 31.5%, restaurants third at 28.0%.

3. ChatGPT 54.0% and Gemini 49.7% were used to choose products and shops (n=300, multiple answers)

We asked the 300 people with experience of using generative AI or AI search for e-commerce which AI they had used to choose products and shops.

4. 45.7% were shown a product, brand or shop they did not know about (n=300)

We asked the 137 people who had been shown something new what action they took (multiple answers). 76.6% looked into it further and 33.6% actually bought it.

Those who bought represent 15.3% of all 300 people using AI for e-commerce.

5. 50.3% abandoned a purchase or switched brand because of an AI answer (n=300)

50.3% in total had one of those experiences. Given the sample size, the 95% confidence interval runs from 44.6% to 56.0%, so reading it as roughly half is reasonable.

6. The most common trigger was a negative statement, at 42.4% (n=151, multiple answers)

We asked the 151 people with that experience what information triggered it. Negative statements about reputation (42.4%), quoted negative reviews (39.1%) and an unfavourable assessment in comparison with other products (33.1%) came top.

All three relate to how AI describes the company. Answers citing information being out of date, or barely any information appearing, together exceeded 30%.

7. 17.0% used the AI answer as it stood; 83.0% checked another source (n=300)

8. Where AI and the official site conflicted: official site 42.0%, generative AI 8.3% (n=300)

The official site scored 42.0%, reviews on the e-commerce site 26.3% and the generative AI answer 8.3%.

9. 38.7% say they consult AI regardless of the amount (n=300)

Excluding the 20.3% who said they do not consult AI, 79.7% intend to consult AI at some price level.

10. 72.0% think companies should understand and improve how AI describes them (n=300)

What the results suggest

The survey shows generative AI working in both directions in shopping: adding candidates and removing them. 45.7% were shown a product they did not know about, and 33.6% of those went on to buy. Meanwhile about half have abandoned a purchase or switched brand because of an AI answer.

The triggers for abandoning or switching are led by negative statements, quoted reviews and unfavourable assessments in comparison. How AI describes a company or product is bearing on the purchase decision.

At the same time, only 17.0% used the AI answer as their basis as it stood; 83.0% checked the official site or reviews on the e-commerce site. Where AI and the official site conflicted, 42.0% trusted the official site. The AI answer does not simply replace the final judgement — it appears to function at the stage of narrowing the candidates.

From these results, we believe companies selling products need to:

• Check regularly how generative AI answers about their products and brands
• Inspect those answers for incorrect or out-of-date information
• Understand which review sites and comparison articles AI refers to
• Keep the basics on the official site — price, specification, stock — accurate, so a consumer arriving to check an AI answer finds the answer there

About the survey

Percentages are rounded to one decimal place, so totals may not come to 100%.
Sampling error is roughly ±2.2 points at n=2,000 and ±5.7 points at n=300 (at a 50% response ratio, 95% confidence).
The figures for n=300, n=137 and n=151 are proportions within the population of people who have used generative AI for e-commerce. They do not represent consumers nationwide.

LLMOcheki https://llmocheki.com/