Observed Signal · Sep 4, 2026 · Research · Source: t3n · Impact: 3/5 · Sentiment: Negative
Study Shows AI Software Recommendations Are Vulnerable to Manipulation
A study by Trellner Research analyzed the sources behind AI recommendations for software, specifically testing Perplexity's Sonar and Sonar Pro models across 380 software categories. The findings reveal that over 80% of the cited source links point to obscure websites, many of which appear designed to be read by AI rather than humans. These sites use tactics like HTML titles such as 'Facts & Grounding Page' and feature hundreds of thousands of generated buying guides, indicating a coordinated effort to manipulate AI search results. The study highlights the growing vulnerability of AI-powered recommendations to SEO manipulation and underscores the need for users to critically evaluate the sources behind AI suggestions.
Reveals a significant vulnerability in AI-driven content recommendations, which could impact advertisers and marketers relying on AI for product recommendations and content strategies.
Track Perplexity Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Trellner Research tested Perplexity's Sonar and Sonar Pro models in 380 software categories.
- 59.8% of cited sources were listed beyond the top 100,000 websites in the Tranco list; 23.4% were not in the top 1 million.
- Three websites with over 215,000 generated buying guides were identified, with sitemaps listing over 100,000 URLs each.
- They shared identical templates and were suspected to be controlled by the same entity, targeting AI retrieval with titles like 'Facts & Grounding Page'.
- The study is limited to Perplexity and does not generalize to other AI models.
Connected Companies & Entities
1 Entity mapped“The AI company whose models Sonar and Sonar Pro were tested in the study....”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Study: AI Recommendations Often Omit Advertorial Labels
A Datapulse Research study found that AI search assistants (ChatGPT, Google AI Overview and Perplexity) frequently cite publisher pages that visibly label content as commercial (advertorials, sponsored content or affiliate pages) without carrying those commercial labels into the AI responses. Datapulse tested 6,668 product/comparison prompts across 120 categories using the Buzzview platform, extracting and analysing over 132k German-language and 139k English-language URLs. In the German dataset 29% of cited sources carried at least one visible commercial marker; Perplexity showed the highest share (31.6%), followed by ChatGPT (28%) and Google AI Overview (26.6%). The study frames the issue as a transparency gap rather than proof of legal wrongdoing.
Shoppers burned by bad AI recommendations, survey finds
A SmartCustomer survey of nearly 1,200 US consumers reveals that while 89% don't fully trust AI recommendations, 76% have used AI to make purchases in the past year, and a third made regrettable purchases based on AI advice. Electronics, apparel, and footwear dominate these poor choices. The survey also indicates that 98% believe AI companies hold responsibility for validating sources and legitimacy, and 53% suspect companies of manipulating AI recommendations with false information. Additional findings show gender and generational divides in AI shopping adoption, with men shopping via AI more than women, and Gen Z being more hesitant. Other surveys by YouGov, RTB House, and Gartner corroborate trust issues and consumer preference for human oversight in AI-assisted commerce.
AI Citation Rankings Often Mislead
The article argues that AI citation rankings are methodologically fragile and often strategically misleading. Rankings depend on a defined set of prompts, and small changes in phrasing can lead to different sources being cited; identical prompts can yield different results over time due to stochasticity in models like ChatGPT, Gemini, Claude, and Perplexity. Different AI systems rely on different data foundations and real-time grounding sources, while training data composition is not publicly disclosed and may overrepresent certain outlets. Grounding sources and training data interact in complex ways, making single-system analyses a poor proxy for overall AI visibility. Since February 2026, Bing Webmaster Tools has begun providing an AI Performance Dashboard showing how often a site’s content is cited in AI-generated answers across Copilot, Bing summaries, and partner integrations, illustrating fragmented visibility data. Google and ChatGPT currently offer no comparable metrics. The piece concludes with four practical approaches to measure AI visibility: focus on topic-specific sources, implement prompt monitoring, conduct brand- and topic-specific tests, and perform cross-system analysis.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.