What happens when you can’t tell who, or what, is behind a recommendation? The internet has always relied on trust signals to help people make decisions. Reviews, discussions, expert recommendations, and user-generated content help consumers determine what products to buy, what brands to trust, and where to spend their money.
As AI-powered search and shopping experiences become more common, those same signals are also influencing machines. AI answer engines, like Google, constantly pull information from sources like Reddit, review sites, and community forums to generate recommendations. As AI becomes another participant in the buying journey, the quality of its recommendations depends on the quality of the information ecosystem it’s built on.
AI Relies on the Information at Hand
The difficult thing about LLMs is that they don’t truly think. They can understand and respond to a query, but unlike humans, they don’t pause and critically think about the question and the information they’ll use to answer it. Their responses are generated from patterns learned across huge amounts of training data, much of which is drawn from content found on the internet. They pull from the types of sources people rely on every day and use that information to determine the most likely answer to a query.
As AI becomes a larger part of the shopping journey, the signals it finds online increasingly influence both consumer decisions and AI-generated recommendations. The challenge is that LLMs can sound incredibly confident, even when the information they’re passing along is completely inaccurate. They don’t fact-check their responses. They identify patterns and probabilities based on the content available to them, which is often biased, misleading, and designed to artificially influence AI outputs.
This creates new challenges for the information ecosystem AI depends on. When misleading, outdated, or low-quality content spreads across the web, AI can absorb and repeat those same patterns. This is similar to the early days of SEO, when keyword stuffing and link spamming were used to game search rankings before search engines became more sophisticated at identifying quality and authority.
The Growing Challenge of Trust Signals
AI is only as trustworthy as the information it’s built on, and people are beginning to test the limits of that trust. In 2024, Google and Reddit announced a partnership that allows Google to use Reddit content to train its AI models. A big reason Google wanted this deal was to help train its models on authentic human experiences and conversations to help their AI models respond in a more natural, human-like way.
The challenge is that AI systems tend to struggle to distinguish genuine expertise from content that seems to be credible. A study done by Cornell Tech revealed that deep-research agents could be influenced through minor edits to public, user-generated content on platforms like Medium, Reddit, and Wikipedia. In one example, researchers demonstrated this by editing a Medium blog post promoting a fake cryptocurrency called BananaCoin. This single misleading sentence resulted in the AI recommending BananaCoin as a promising crypto investment, despite it not existing. Those seemingly minor changes were enough to influence how AI research agents evaluated and recommended information, highlighting how susceptible these systems can be to manipulated trust signals.
The issue extends beyond intentional manipulation. The rapid growth of AI-generated content, or “AI slop” is flooding the internet with low-quality articles, fake reviews, and mass-produced content designed to attract clicks or influence search visibility. Even when content isn’t intentionally deceptive, it adds more noise to an information ecosystem that AI increasingly depends on, making it hard to separate credible information from unreliable sources. As AI systems increasingly encounter AI-generated content, that noise can compound, increasing the risk of hallucinations and creating a feedback loop where more content produces progressively less useful information.
The BBC demonstrated just how easily these signals can influence AI-generated answers. In one experiment, a journalist was able to convince Google and ChatGPT that he was a champion competitive hot dog eater simply by publishing a single article on his personal website. The example is silly and harmless, but it raises a more important question– if AI can be misled about something so trivial, what happens when the subject is financial advice, healthcare information, or choosing a business to work with? That risk becomes more tangible when companies have a financial incentive to shape the information AI encounters. Shopify, for example, has published at least 60 articles comparing ecommerce platforms, placing itself first in each. When The Atlantic tested the effect, ChatGPT recommended Shopify as the best way to set up an online storefront and cited Shopify’s own comparison pages as sources.
As more people turn to AI-generated summaries instead of evaluating multiple sources themselves, the quality of the underlying information matters now more than ever. In this environment, trust is no longer a given. It’s something that will have to be continually earned and reinforced.
Trust is Becoming a Competitive Advantage
For commerce brands, this changes the conversation. Product pages, customer reviews, buying guides, FAQs, and support documentation provide valuable signals that help AI systems understand products and recommend brands. While companies can’t control conversations happening online, they can strengthen the information they do control by keeping product data accurate, demonstrating expertise, and creating authoritative first-party content.
There’s a technical foundation that also matters. Structured data, crawlability, accessibility, and clear content organization help search systems discover and interpret information. But technical accessibility alone does not establish credibility. Google’s guidance for AI Search continues to emphasize the fundamentals of helpful, reliable, people-first content, while its recent updates have started to focus on helping users find original content, firsthand perspectives, and sources they value.
Google introduced Preferred Sources to AI Overviews and AI Mode, allowing users to identify sites they want to see more prominently in AI-generated responses. Google says people are twice as likely to click through to a Preferred Source. Google also introduced a “Highly Cited” label to help users identify original reporting and influential coverage. These features suggest a broader effort to give greater visibility to sources with established value and original contributions, rather than treating every piece of content as an equally useful signal.
That matters because trust is increasingly built across multiple, reinforcing signals. Brands can’t optimize a single page or add a single technical feature and expect it to establish credibility. The goal is to build an information ecosystem where product claims, structured data, customer experiences, expert perspectives, and third-party sources consistently reinforce one another.
The examples mentioned earlier show how quickly perceived AI ranking signals can become shortcuts, with companies creating content designed to shape how AI systems perceive and recommend them. As these tactics become more visible, search and AI platforms have an incentive to put stronger guardrails around how information is surfaced. Google explicitly says its systems are designed to prioritize helpful, reliable information created for people rather than content created to manipulate search rankings, and its spam policies now explicitly address attempts to manipulate generative AI responses.
For brands, that makes trust a broader strategic consideration. Accurate data, credible first-party content, authentic customer signals, independent validation, and consistent information across channels can create a stronger foundation for discovery and conversion across both traditional and AI-powered experiences. As AI becomes a larger part of the buying journey, trust becomes an increasingly valuable business asset. In a digital landscape where information is becoming easier to create and hard to evaluate, credibility might be one of the few advantages that gets harder to manufacture at scale.