AI Models Have Shown Inconsistent Shopping Results

New research found that top AI models frequently change product recommendations and cite differing sources.

Updated on Sept. 30, 2026 in Artificial Intelligence

Isometric editorial illustration of shifting ceramic shopping bags, representing the inconsistency of AI-generated shopping recommendations. — Recent research from Suff Digital and MentionBird shows that top AI models frequently provide inconsistent and shifting product recommendations during repeat shopping queries. AI Illustration. Upload story photo >

Live Poll

Do you trust AI-generated product recommendations for your own shopping and purchasing decisions?

Recent studies from Suff Digital, Product.ai, and MentionBird indicate that ChatGPT and Gemini often provide conflicting product recommendations. These platforms frequently change their top choices when asked the same question multiple times in back-to-back iterations.

Why it matters

The variability in AI-generated shopping advice complicates how consumers rely on these tools for purchasing decisions. Frequent changes in top recommendations and inconsistent sourcing suggest a lack of reliability in current AI models for commercial research.

Product.ai found 86% of 220 tested shopping questions produced factual conflicts across models. Additionally, an academic study noted that ChatGPT and Gemini shared no source domains in 76.7% of comparisons.

The players

ChatGPT

This is a generative artificial intelligence chatbot developed by OpenAI that is widely used for information retrieval and content generation.

Gemini

This is a suite of generative artificial intelligence models developed by Google that integrates into search and product-related assistance.

Suff Digital

This research firm focuses on analyzing the behavioral and output patterns of generative artificial intelligence models.

Product.ai

This organization specializes in evaluating the accuracy and reliability of AI systems when applied to consumer shopping inquiries.

MentionBird

This analytics company provides tracking data on digital trends and the performance of AI-driven recommendation engines.

The details

Suff Digital asked 300 shopping questions 10 times each, finding high rates of recommendation churn. Meanwhile, MentionBird analyzed over 77,000 total answers to track shifts in top-five brand lists during consecutive model transitions.

August 2026: Tracking showed Reddit citations in ChatGPT fell.
August 27, 2026: MentionBird published study results on recommendation churn.
September 16, 2026: Academic preprint published regarding source variation.
September 22, 2026: Product.ai published study results.
September 24, 2026: Suff Digital published study results.

The Tech Race

These findings reflect a broader trend where AI models reduce reliance on specific user-generated content platforms like Reddit, as noted in August 2026. This repositioning highlights the ongoing struggle to stabilize recommendation accuracy in a competitive and rapidly evolving tech sector.

Users should be aware that asking the same shopping question multiple times may yield drastically different results. Relying solely on a single AI response for expensive or high-stakes purchases could lead to inconsistent or conflicting brand recommendations.

The takeaway

Consumers should verify AI-generated product suggestions with independent retail sources rather than treating the chatbot as a definitive expert. Always re-run important queries to see if the model produces stable results across multiple attempts.