Two of the world’s most dominant technology companies—Google and Microsoft—are now making a bold claim: artificial intelligence is ready not only to change but to actively improve the way we browse and interact with the information-rich labyrinth of the modern web. This week witnessed the debut of Google AI Mode in the UK, following successful launches in the US and India, as well as the release of Microsoft’s Copilot Mode integrated directly into Bing’s ubiquitous search bar. Both offerings promise seismic shifts in search, integrating conversational capabilities directly into the browsing experience and touting the apparent end of static keyword queries. But do these highly polished AI search bars actually represent the evolution that users have been promised, or are they well-marketed, incremental upgrades to systems that still fundamentally rely on traditional engine mechanics? Through rigorous real-world testing, coupled with cross-verification from independent expert analysis, this feature explores how these AI-infused search engines stack up—and whether they truly deliver on their ambitious claims.
Setting the Stage: What Exactly Are Google AI Mode and Copilot Mode?
Google’s new AI Mode is built around Gemini 2.5, the company’s multimodal language model that synthesizes results across text, speech, and visual queries. Users can type, speak, or even upload photos to ask questions ranging from “recommend an e-bike for city commuting” to “how do I change a flat tire?”—with Google advertising capabilities that span product comparison, trip planning, and breakdowns of complex instructions.Microsoft’s Copilot Mode in Bing, on the other hand, pushes a similar narrative with its Copilot AI model. According to the company, Copilot “reads, compiles, and reasons about information available on the web,” then presents findings in the form of concise, synthesized answers. Both platforms position themselves not as mere search engines but as proactive digital partners—AI co-pilots that help users make sense of tangled and contradictory internet knowledge.
From Search Bar to Chatbot: The Promise of Conversational Queries
Testing conducted by the Euronews Next team, as well as additional evaluations from WindowsForum experts, paints a clear picture: both Google AI Mode and Copilot Mode essentially function as conversational chatbots within the search bar. Type in a multi-step question (“generate a week-long meal plan with 1,800 daily calories, no dairy, with shopping lists”), and both AI search engines return human-readable, structured responses. Ask for e-bike recommendations, and you might get a breakdown with pros, cons, and links to online stores—sometimes even tailored to the specific use-case you hint at in your query.Notably, both AIs are proactive in inferring user intent, often making assumptions about user needs. In one test, both platforms crafted a meal plan with the implicit assumption that the user had some cooking experience and no additional dietary restrictions, despite those not being explicitly stated. This shows a leap in sophistication and context-awareness compared to legacy search, but also introduces risks: when an AI makes silent assumptions, the utility may be high for some users, but misleading or even unsafe for others (especially in situations involving allergies, accessibility needs, or medical queries).
Copilot’s “Context Clues” and Living-Browser Integration
A standout differentiator between Microsoft’s and Google’s approaches is Copilot’s leveraging of “context clues” from user browser tabs. Enable this in Copilot, and the search AI will actively adjust its recommendations and answers based on the context set by your open tabs.For instance:
- With no tabs open, Copilot recommended open-ended job listings from aggregation sites.
- With LinkedIn, Glassdoor, Indeed, and Reed open, Copilot narrowed its responses, pulling job posts exclusively from these sites.
Critically, Google AI Mode takes a more walled-garden approach: it answers queries based solely on the prompt—with no tab-based contextualization. This may mean less tailored results in some cases but reduces the potential for privacy pitfalls arising from AI overfitting to a user’s immediate digital environment.
Addressing Controversial Topics and Misinformation: A Measured, Discursive Approach
Perhaps the greatest challenge for AI search is not answering simple factual questions, but synthesizing disputed or controversial topics rife with misinformation. To test this, both engines were prompted to summarize diverse viewpoints about the measles vaccine—a topic notorious for being beset by strong consensus, but also persistent fringe skepticism and outright falsehoods proliferating online.Both Copilot and Google AI Mode foregrounded the scientific consensus: that measles vaccines are safe, highly effective, and critical to public health, while acknowledging extremely rare side effects. When asked for sources, Copilot promised reliance on high-quality references, specifically naming the World Health Organization, the US Centers for Disease Control (CDC), and peer-reviewed literature.
Interestingly, if browser tabs containing known misinformation sites were open during the search, Copilot’s responses shifted slightly. While the scientific consensus remained central, the model surfaced more skeptical or “alternative” viewpoints, explaining that it was doing so because “these perspectives should be contextualized carefully to avoid amplifying misinformation.” Google AI Mode, meanwhile, included dissenting perspectives with a clear caveat: “It’s essential to include this viewpoint because it represents a significant part of the public discourse, even if it contradicts the scientific consensus.”
This nuanced discursive style is an improvement over blunt fact regurgitation, but it is inherently risky: too aggressive an inclusion of fringe views risks false equivalence, while their exclusion opens the AI to accusations of bias or censorship. Neither Microsoft nor Google would provide details on exactly how their models judge what is a “high-quality source,” a lack of transparency that, as multiple digital ethicists note, undermines full trust in the results.
The Limits: Task Handling, Concierge Features, and Unmet Promises
Despite promotional videos hinting at “digital concierge” abilities, today’s AI search bars remain, fundamentally, search engines—albeit more sophisticated and context-aware than ever. In testing, Google AI was able to generate meal plans, but could not actually source ingredients or add them to a shopping cart. Copilot, which Microsoft manuals once showed booking a paddleboard rental, could not actually make dinner reservations, book medical appointments, or upload CVs to job boards in real-world tests. To queries requesting such services, both AIs replied with polite rejections or referred users to “future updates.”There is a broader pattern here, mirrored across both AIs:
- Generating information summaries, guides, and product comparisons: Highly capable
- Integrating with third-party services for task automation and purchasing: Not yet realized
- Handling complex, multi-step tasks autonomously: Still mostly aspirational
User Experience and Interface: Breaking the Query Paradigm
The conversational aspects of both search bars represent a marked shift from previous paradigms. Instead of typing lists of disconnected keywords, users now engage in fluid “conversations” with the AI: asking follow-up questions, requesting clarifications, or steering the AI with additional instructions. For casual users and information hobbyists, this feels both welcoming and vastly more interactive.However, power users and professionals may find limitations. The AI may still “hallucinate” details, omit critical citations, or fail to provide explicit links to raw sources—issues that can hamper rigorous fact-checking or academic research. Testing reveals that while both engines are generally diligent in sourcing reputable agencies like the WHO or CDC for health information, neither consistently provides live, clickable citations for every factual claim within their summaries, making source auditing more cumbersome.
Privacy Trade-offs: A Double-Edged Sword
Copilot’s context clue feature stands as both triumph and lightning rod. The ability for the AI to dynamically personalize responses based on open tabs is a breakthrough in contextual relevance—it bridges the gap between passive search engines and truly interactive digital assistants. Yet it is also fraught with privacy implications. What else is the AI tracking? How long is this data stored? Are users aware of when this contextualization is active?While Microsoft touts this as an opt-in feature, the company’s silence regarding technical safeguards and user controls leaves lingering questions. Google’s approach, by not observing open tabs, is arguably more privacy-preserving, but also risks delivering information that’s not as finely attuned to the human on the other side of the screen.
Critical Analysis: Are We Seeing Revolution or Incrementalism?
At their core, both Google AI Mode and Copilot Mode represent a significant evolution—if not a revolution—in how we search and interact with information online.Strengths:
- Conversational engagement lowers the barrier for complex queries.
- Contextual awareness (especially in Copilot) provides more tailored, dynamic answers.
- Explicit handling of misinformation and controversial topics signals a maturing of responsible AI use.
- Multimodal input (especially in Google’s case, with voice and images) extends utility beyond standard text.
- Assumptions about user intent can yield results that are spookily accurate or wildly off-base, depending on the individual.
- Transparency in source attribution is inconsistent; neither system provides fully auditable sourcing on-demand.
- Actual automation—handling appointments, purchases, uploads—remains largely theoretical, not operational.
- Privacy and data protection, especially in contextualization features, remain under-described and under-scrutinized.
Both platforms demonstrate greater nuance—but also greater risk—in presenting fringe or controversial perspectives. While intended to responsibly contextualize, not amplify, fringe beliefs, the implementation must be constantly monitored and tuned to avoid the unintentional spread of harmful or unscientific information.
Comparison Table: Google AI Mode vs. Microsoft Copilot Mode
| Feature | Google AI Mode | Microsoft Copilot Mode |
|---|---|---|
| Model Core | Gemini 2.5 AI | Copilot / GPT-infused model |
| Multimodal Input | Text, voice, images | Primarily text; some image support |
| Context Awareness | Query only | Uses browser tabs for context (opt-in) |
| Task Automation | No—search only | No—search only |
| Misinformation Handling | Highlights consensus, caveats for alternatives | Contextualizes based on scientific consensus |
| Source Attribution | Some, mainly established agencies | Claims high-quality sources, limited detail |
| Privacy Safeguards | No browsing context tracked | Context feature raises privacy questions |
| Availability | US, India, UK | Global (varies by Bing region) |
User Implications: What Should You Expect?
For most mainstream users, these AI search bars yield a more engaging, user-friendly search experience, boosting productivity for exploratory queries, product research, and nuanced questions. The integration of conversational, context-aware AI into the very fabric of search will likely raise expectations for all search engines in the coming years.Power users expecting seamless integration into their workflows—uploading documents, auto-filling forms, or delegating tasks—should temper expectations in the immediate term.
Privacy-conscious users should be especially vigilant. The use of context cues, while powerful, is largely opaque and ripe for misuse if unchecked by regulatory and consumer pressure.
Conclusion: A First Step Toward the Future of Search
Microsoft and Google’s new AI modes do not yet represent the autonomous, action-oriented web companions of science fiction. Instead, they are a major step in the march toward that goal—turning search engines into something more like conversation partners who can reason about intent, provide syntheses, and contextualize a world brimming with nuance, contradiction, and misinformation.For now, users will find these search bars both more useful—and, in some respects, more unpredictable—than what came before. The true test will be whether these platforms fulfill their promises to integrate meaningful automation, transparent sourcing, and robust privacy protections. Until then, both Google AI Mode and Copilot Mode serve as a compelling glimpse into the next era of search—one characterized by conversational intelligence, contextual sensitivity, and the ongoing negotiation between utility and trust.
Source: inkl AI companies want to browse the web for you. We tested 2 to see how well they actually work
