Google Search is undergoing its most dramatic transformation in a quarter-century, driven by the mainstream deployment of Gemini AI. Here’s a breakdown of the key developments.
The search box redesign
For the first time in 25 years, Google redesigned its search bar in May 2026. The new interface is larger and more interactive, built to handle longer, more complex, multi-part queries rather than simple keyword terms. Users can now include images and videos directly in their searches, and a chatbot is integrated into the main search interface for follow-up questions.
AI Mode becomes the default
Google’s AI Mode (powered by Gemini 3) is now the default search experience. It moves away from keyword matching and toward interpreting intent, handling multi-layered questions by considering all constraints and returning visually structured responses alongside links to source pages. Gemini now accounts for more than a quarter of all generative AI traffic, up from just 7% a year ago.
Agentic features
For Google, search is no longer just about finding information it can now act on it. Google has introduced digital “agents” into Search, meaning a user looking for a flat could receive notifications about new listings without ever visiting Rightmove or Zoopla. This agentic capability, derived from Google’s Project Mariner, is also being built into Chrome, where Gemini can execute tasks like adding items to a basket on a user’s behalf.
This is not helpful for those of use who use Google primarily for due diligence work.
Gemini embedded in Chrome
Beyond Search, Gemini has been embedded directly into Chrome, now used by nearly 3 billion people worldwide. It can synthesise information across multiple open tabs, interpret webpage content, and pull data from Gmail to build itineraries or answer queries. By the end of 2025, Google also integrated AI Mode into Chrome’s Omnibox (address bar), making Gemini effectively the default interface for the world’s most dominant browser.
Concerns and criticisms
Critics have flagged that Gemini in Search, like so many AI platforms sometimes fabricates answers generating responses that appear relevant but lack accuracy or contextual grounding. There are also broader concerns that traditional blue-link search is dying, with AI summaries reducing clicks to source websites, this is a significant issue for publishers and, notably for due diligence investigations as source verification is crucial.
What this means for online research & Due Diligence work
For investigative and compliance work, the shift is significant. AI-generated overviews synthesise rather than cite, which can obscure the provenance of information. As Gemini interprets queries rather than returning raw results, it becomes harder to surface the niche, non-mainstream, or adverse information that a well-constructed keyword search would previously have captured. Relying on AI Mode alone for subject searches risks missing critical red flags buried in less-indexed sources.
What appears to still work (for now anyway)
Key operators that remain functional in 2026 include:
filetype:pdf, filetype:docx, filetype:ppt (still working for document types, phew!)
site: (restrict results to a specific domain) a really useful search tool
“exact phrase” in quotes, technically still works, but may be ignored by Google in favour of AI results
intitle:, inurl:, intext: (still supported)
before: and after: (date range filtering, although I would still recommend using the tools drop down)
OR, AND, – (minus to exclude terms)
Implications
For investigators, this is a real degradation. The technique commonly known as Dorking using stacked operators to surface hidden documents, leaked spreadsheets, or specific filetypes on specific domains is becoming less reliable as AI Mode ‘interprets’ rather than executes your query.
Platforms like Maxintel and dedicated OSINT references now publish cross-engine compatibility tables specifically because Google’s operator support has become inconsistent enough that investigators are routing searches through alternative engines like Bing, DuckDuckGo, or Yandex where operators remain more dependably executed.
The practical takeaway. Google is not formally announcing the death of advanced search operators, but it is engineering a system where they matter less and less.
For rigorous investigative research, you would be wise to treat Google operators as a best-effort tool and cross-reference with engines/databases and dedicated websites.
AI is good for speed, it is not so good for in-depth investigative work, we have implemented AI into our EDD-Pro Technology, but not at the expense of other more conventional tools. Head over here to find out more.
If you got this far I have a question, did you spot the irony?
