Google’s advanced search has changed in three major ways over the last few years:

1. It has become far more AI-driven and conversational.
2. Many traditional operator-based workflows are less visible or less reliable.
3. Search results are increasingly synthesised instead of simply ranked.

Here’s the practical breakdown.

1. Google moved from “query matching” to “answer generation”

Historically, advanced search meant:

• Boolean operators (AND, OR, -)
• Exact-match quotes
• site:
• filetype:
• intitle:
• date filtering
• highly structured keyword queries

Google still supports many of these, but the core UX is no longer centred around precision operators. Instead, Google now pushes:

• natural-language questions
• follow-up conversational search
• multimodal search (voice, image, screenshots)
• AI-generated summaries (“AI Overviews”)
• “AI Mode” conversational results

Example: Old-style advanced query: site:gov.uk “money laundering” filetype:pdf

New-style Google behaviour: What are the current UK anti-money laundering guidance documents for regulated entities? Google increasingly tries to infer:

• intent
• context
• entity relationships
• authority
• likely next questions

instead of strictly obeying keyword syntax.

2. AI Overviews changed the structure of search results

Google now frequently inserts AI-generated summaries at the top of results pages. These:

• synthesise information from multiple sources
• answer informational queries directly
• reduce the need to click on websites
• often appear before traditional blue links

This fundamentally changes advanced search because:

• users receive interpreted answers instead of raw source lists
• source selection is less transparent
• ranking signals are no longer the whole story

Research in 2026 found:

• AI Overviews appear for a large percentage of informational searches
• cited sources can differ substantially from standard organic rankings
• responses can vary between runs of the same query

So advanced searching is shifting from: “find documents” to: “Interrogate a synthesis engine”

3. Some classic operators are weaker or inconsistently supported

Many advanced operators still work:

• site:
• filetype:
• quotes ” “
• -keyword
• OR
• before: / after:

But Google has gradually reduced emphasis on:

• highly granular operator documentation
• exact matching
• predictable Boolean logic

Certain legacy operators have been deprecated or have become unreliable over time, including:

• link:
• info:
• some uses of inanchor:
• precise wildcard behaviour

Modern Google prioritises semantic relevance over literal operator precision.

4. Search became multimodal

Advanced search now includes:

• image search via Google Lens
• “Circle to Search”
• voice queries
• screenshot-based lookup
• camera-driven search workflows

Google reports massive growth in visual search usage. That means “advanced search” increasingly includes:

• visual context
• object recognition
• OCR
• scene understanding
• geospatial inference

5. “Web-only” search is now a niche workflow

A growing number of users deliberately bypass AI-enhanced results using:

• Google “Web” tab
• udm=14 URL parameter

This strips out:

• AI Overviews
• shopping modules
• many rich SERP features

and restores a more traditional link-based experience.

Example: https://www.google.com/search?q=osint&udm=14

This has become popular among:

• OSINT practitioners
• researchers
• journalists
• technical users
• archivists

who want less interpretation and more direct visibility into the source.

6. SEO and discoverability changed dramatically

Google’s newer search systems increasingly reward:

• authority
• experience-led content
• structured data
• citations
• topical trust
• entity relationships

rather than pure keyword optimisation. For advanced search users, this means:

• fewer obscure niche sources surface organically
• high-authority domains dominate AI summaries
• long-tail discoverability is reduced in some contexts

7. Advanced search is now partly “prompt engineering”

Experienced users increasingly search using:

• long natural-language prompts
• layered contextual questions
• iterative refinement
• conversational follow-ups

Example progression: Find Isle of Man AML regulations for TCSPs
then: Only official regulatory guidance
then: Compare with FATF Recommendation 10

This resembles interacting with an LLM more than with classic search engine syntax.

What has not changed

Traditional advanced operators remain extremely useful for:

• OSINT
• compliance research
• legal/regulatory discovery
• academic work
• cyber investigations
• metadata hunting

Especially:

site:
filetype:
“exact phrase”
before:
after:

These are still among the highest-value precision tools in Google Search.

Net effect

Google’s advanced search has evolved from:

For technical users, investigators, analysts and researchers, the biggest adaptation is:

• combining classic operators with conversational querying
• explicitly forcing web-only results when necessary
• validating AI summaries against primary sources rather than trusting synthesised output unquestioningly.

Quotes (” “) and asterisks (*) used to be central to Google’s advanced search precision, but their behaviour has changed noticeably as Google shifted toward semantic interpretation and AI-assisted retrieval.

Quotes “exact phrase”

Historically:

• Quotes forced near-literal phrase matching.
• Google returned pages containing the exact word sequence.

Example: “enhanced due diligence”
Originally, this meant:

• all three words
• in that order
• adjacent or nearly adjacent

What changed

Google still treats quotes as a strong precision signal, but:

• semantic expansion sometimes still occurs
• stemming/pluralisation can leak in
• Snippets may highlight non-exact matches
• AI Overviews may summarise beyond the exact quoted text

However, quotes remain one of the most reliable operators for:

• legal text
• OSINT
• plagiarism checking
• identifying copied content
• finding exact document wording
• verifying whether a phrase actually exists online

Quotes are now relatively more valuable.

Because Google has become more interpretive overall, exact quotes are now one of the few remaining ways to achieve higher precision.

Example: site:iomfsa.im “source of wealth”
This is still highly effective for regulatory discovery.

Asterisk *

The asterisk operator changed much more dramatically.

Original behaviour

Historically: “Tony * Bennett” would act roughly like:
wildcard for one or more unknown words.
Google could match:

• Tony J Bennett
• Tony Alan Bennett
• Tony and Bennett

Similarly: “money laundering * regulations”
might match:

• money laundering reporting regulations
• money laundering prevention regulations

This was extremely useful for:

• partial quotations
• song lyrics
• fragmented text recovery
• OSINT
• document reconstruction

Current behaviour of *

Google no longer documents * as a true wildcard operator in the same way. Today, it behaves inconsistently:

• sometimes ignored
• sometimes treated as a loose placeholder
• sometimes semantically expanded
• often less deterministic than before

In practice:

• It still occasionally works inside quoted phrases
• It is much less reliable for precision searching
• semantic search often overrides literal wildcard intent

Example: “the * fox jumps”
may return:

• exact wildcard-like substitutions
• semantically related phrases
• approximate matches

instead of strict positional substitution.

Modern replacement for wildcard workflows

Advanced users now often use:

• partial quotes
• OR logic
• multiple searches
• AI prompting
• regex-capable external search engines
• specialised OSINT tools

instead of relying on Google * .

Example modern workflow: Instead of: “money laundering * regulations”
users now do:
“money laundering regulations” OR “money laundering reporting regulations” OR: site:gov.uk “money laundering”
then refine iteratively.

Important nuance: Google now “understands” missing words

Because modern Google uses entity understanding and embeddings:

• it often infers omitted words automatically
• wildcard precision matters less to Google internally
• but matters more to investigators/researchers

This is one reason the explicit * operator lost importance.

Google assumes: “I know what the user probably means.”
Whereas traditional advanced search users often want: “Return only literal structures.”
That philosophical shift is a major change in search behaviour.

Current best practices

Use quotes for:

• exact wording
• legal/regulatory language
• leaked-text validation
• document tracing
• name disambiguation
• OSINT verification

Example: “beneficial ownership register”

Use * only experimentally

It can still help occasionally for:

• forgotten song lyrics
• partial quotations
• fragmented text

Example: “to be * not to be”
But do not rely on it operationally.

Better modern alternatives

For precision: “exact phrase” site:domain.com
For variation handling: (“EDD” OR “enhanced due diligence”)
For date scoping: after:2024 before:2026

For investigative work:

• Google + alternative engines
• OCR tools
• vector search
• specialized databases
• AI-assisted summarisation
• direct site search

Bottom line

Quotes remain essential for advanced users.
The wildcard operator is now more of a legacy artefact than a dependable precision tool.

Regex-capable external search engines

Regex-capable or regex-adjacent search tools are typically outside the mainstream of web search engines like Google Search because Google intentionally abstracts away low-level pattern matching. For OSINT, DFIR, compliance, cyber investigations, research and archival work, practitioners instead use platforms that support:

• regular expressions directly
• Lucene syntax
• grep-like pattern matching
• advanced indexing/query DSLs
• proximity/wildcard/fuzzy logic beyond standard search engines

Here are the major categories:

True regex-capable search tools

https://grep.app/

Searches public GitHub repositories with regex support.

Excellent for:

• secrets hunting
• malware research
• API key discovery
• code intelligence
• infrastructure fingerprinting

Example: AKIA[0-9A-Z]{16}
Finds exposed AWS access keys.
Another: password\s*=\s*[“‘]
Finds hardcoded password assignments.

https://sourcegraph.com (Paid)

Enterprise-grade code search with strong regex support.

Supports:

• regex
• structural search
• repository scoping
• symbol indexing

Widely used in:

• AppSec
• large engineering environments
• vulnerability research

https://www.shodan.io

Not a full regex, but it supports powerful banner matching and filtering.

Useful for:

• exposed services
• ICS/SCADA
• internet-facing infrastructure
• TLS/certificate analysis

Example: product:”nginx”
or:
ssl:”example.com”

https://search.censys.io

Supports:

• fielded searching
• certificate parsing
• host enumeration
• protocol metadata analysis

Closer to query DSL than consumer search.

https://www.elastic.co (Paid)

The backbone of many enterprise investigation systems.

Supports:

• Lucene regex
• wildcard
• proximity
• fuzzy matching
• Boolean logic

Example Lucene regex: /error-[0-9]+/
Common in:

• SIEM
• SOC operations
• log analysis
• compliance monitoring

Regex-adjacent investigative platforms

https://www.maltego.com

Not regex-centric, but supports transform-based data correlation.
Used heavily in:

• OSINT
• link analysis
• entity mapping

Often combined with regex extraction workflows.

https://www.intel471.com (Paid)

Can:

• scrape data
• extract indicators
• pattern-match emails/domains/IPs

Useful for automated reconnaissance.

https://github.com/lanmaster53/recon-ng

Often combined with:

• regex pipelines
• scraping
• grep/sed/awk tooling

More technical workflow.

Web search engines with limited advanced syntax

https://search.brave.com

Supports some advanced operators and tends to preserve literal matching better than Google in certain contexts. Still not a true regex.

https://kagi.com

Power-user-oriented search engine with:

• lenses
• prioritisation
• domain weighting
• cleaner literal handling

Again, not regex, but often preferred by researchers frustrated with Google’s semantic drift.

Archive and dataset search tools

https://web.archive.org/

Not regex-native, but useful for:

• historical reconstruction
• deleted content
• comparing snapshots

Often paired with regex locally after export.

Common Crawl

https://commoncrawl.org/

Massive web crawl datasets.

Researchers frequently:

• download indexes
• run regex locally
• perform NLP extraction

Very powerful but technical.

CLI and local regex workflows

For serious investigators, regex search often moves locally.

Core tools include:

• grep
• ripgrep
• awk
• sed
• jq
• yq

Especially: ripgrep
Extremely fast recursive regex search.

Example: rg -i “beneficial owner”
Regex: rg “[A-Z]{2}[0-9]{6}”

Widely used in:

• DFIR
• code auditing
• leaked-data analysis
• SOC workflows

Why Google moved away from regex-style search

Google optimised for:

• consumer usability
• intent inference
• AI summarisation
• semantic embeddings
• conversational search

Regex conflicts with that model because regex assumes:

• deterministic retrieval
• literal structure
• syntax discipline

Google now prioritises:

• probabilistic relevance
• inferred meaning
• blended answers

That makes regex-style workflows increasingly external to mainstream search.

Practical OSINT stack today

A modern investigator might combine:

Most useful regex-capable tools by discipline

https://www.google.com/advanced_search

Google Advanced Search is basically a form-based query builder. The boxes convert your entries into Google operators and URL parameters, so you do not have to type them manually.

Google’s own Advanced Search page describes the main boxes this way: quotes for exact phrases, OR for alternatives, minus signs for excluded terms, two dots for number ranges, site/domain restriction, location of terms on the page, file-type filtering and usage-rights filtering.

For OSINT, the most useful boxes are usually:

All these words: fraud investigation
Exact phrase: “Tony Bennett”
Any of these words: leak OR breach OR database OR dump
None of these words: music singer
Site or domain: gov.uk
File type: pdf
Terms appearing: in the title of the page