Safeguarding Business Data: Strategies to Mitigate AI Web Search Accuracy Risks

Over half of us have embraced AI as a primary tool for web searches, yet many still grapple with concerns over data reliability. The rise of generative AI (GenAI) undoubtedly offers remarkable efficiency, but recent investigations reveal a troubling gap between user confidence and the actual accuracy provided by these platforms. For businesses, this misalignment could lead to serious repercussions in compliance, legal standing, and financial planning.

The Trust Discrepancy in AI Tools

For executives, the integration of AI tools presents a classic “shadow IT” dilemma. A survey conducted in September 2025 found that approximately one-third of respondents view AI as more integral to their daily lives than traditional web searching. If employees are placing their trust in these tools for personal inquiries, it stands to reason they are also relying on them for business-related research.

The investigation from Which? underscores that unverified trust can be detrimental. Close to half of AI users express a high level of confidence in the information generated. However, upon closer examination, such trust may be unjustified.

The Accuracy Gap: What the Study Reveals

The study evaluated six popular AI tools—including ChatGPT, Google Gemini, Microsoft Copilot, Meta AI, and Perplexity—across 40 common questions related to finance, law, and consumer rights.

• Perplexity topped the list with a score of 71%.
• Google Gemini AI Overviews followed closely at 70%.
• Meta AI lagged behind at 55%.
• ChatGPT, widely used yet underperforming in this assessment, scored 64%.

This disparity shows that popularity doesn’t always equate to reliability.

Despite their advanced technology, these AI tools often misinterpret information or provide incomplete advice that could jeopardize business integrity. For finance and legal departments, the nature of these inaccuracies could be particularly concerning.

For instance, when asked about investing a £25,000 annual ISA allowance, both ChatGPT and Copilot overlooked a crucial error regarding statutory limits. Instead of correcting the mistake, they provided advice that could lead to violations of HMRC rules.

Although platforms like Gemini, Meta, and Perplexity successfully identified such errors, the inconsistencies underscore the necessity for a “human-in-the-loop” approach in any business context utilizing AI technology.

Legal teams also face challenges; AI’s tendency to generalize regulations can create significant risks. The study found that AI models frequently overlook the fact that legal statutes vary significantly between UK regions. This is especially applicable when considering nuances between Scotland and the rest of the UK.

Moreover, the study brought to light an ethical dilemma: AI tools rarely advise users to consult registered professionals for high-stakes queries. For example, when prompted about a dispute with a contractor, Gemini suggested withholding payment, potentially placing the user in breach of contract.

This “overconfident advice” could result in operational hazards for businesses relying on AI for crucial compliance checks or contract reviews without proper verification.

Source Transparency and Data Governance

One major concern for data governance within enterprises is the integrity of the information being utilized. The investigation indicated that AI tools often face challenges with source transparency, citing vague or downright unreliable sources, such as outdated forum discussions. This lack of clarity can lead to unnecessary financial burdens.

In one instance regarding tax codes, both ChatGPT and Perplexity directed users to premium tax-refund services instead of the official HMRC resource, potentially incurring high fees for users.

In a business procurement scenario, algorithmic biases could lead companies to overspend or engage with vendors that do not meet requisite corporate standards. Major tech companies recognize these limitations, often placing responsibility for verification squarely on the user—and by extension, the enterprise.

A spokesperson from Microsoft clarified their tool’s role as a synthesizer rather than an authoritative source, urging users to verify content accuracy.

OpenAI, addressing these findings, asserted that improving reliability is a collective industry effort, noting that the upcoming GPT-5 model promises enhanced performance.

Strategies for Mitigating AI Business Risk

For business leaders, the objective should not be to eliminate AI tools, as this can push their use further into the shadows. Instead, the focus should be on establishing robust governance frameworks to ensure their accuracy in web searches:

1. Ask Specific Questions: AI is still in its learning phase. Employees should frame precise queries to avoid risk-laden data results. For instance, specifying “legal rules for England and Wales” helps ensure the AI understands the jurisdictional context.

2. Verify Sources: Relying on a single output is a poor practice. Employees should always verify information across multiple AI platforms or seek a “double source” for critical topics.
3. Encourage a Second Opinion: At this point in GenAI development, it’s essential that AI outputs are treated as one of multiple opinions. For complex issues involving finance, law, or healthcare, human expertise must ultimately guide decision-making.

While AI tools are progressing and their accuracy is gradually improving, the investigation concludes that over-reliance on them at this stage could lead to costly mistakes. For enterprises, finding the balance between AI-driven efficiency and responsible compliance lies in a diligent verification process.

As you navigate the evolving landscape of AI, remember: informed decisions come from a combination of technology and human insight. Embrace the journey with awareness and caution, and let’s harness the power of AI responsibly.