Last updated: May 20, 2026

Google Search is changing from a list of websites into an answer machine. That sounds convenient until you look at what it means for privacy, publishers, and user choice.
AI Overviews can summarize information before you ever click a source. Google says this helps people ask more complex questions and find information faster. The problem is that users, publishers, and independent website owners are being pushed into a new bargain they did not clearly choose: Google gets the query, Google gets the answer box, Google keeps the user, and the open web gets whatever clicks are left over.
Contents
- Quick answer
- What Google AI Search Changes
- Why People Are Angry About AI Overviews
- The Click Problem: Sources Get Cited, But Do They Get Traffic?
- What This Means For The Internet
- Why You Should Care Even If You Like AI
- Best Privacy Search Alternatives
- How To Run SearXNG On A VPS
- Reverse Proxy With Nginx
- Can You Run SearXNG On Android With Termux?
- How To Make SearXNG Your Browser Search Engine
- Final Recommendation
- Follow The Thrifty Dev
- FAQ
- Citations and Further Reading
- Related Posts
- Keep building from here
Quick answer
If you want search results without being nudged through Google’s AI answer layer, try one of these:
- SearXNG: Best if you want self-hosted metasearch and control.
- Brave Search: Best easy default replacement with an independent index.
- Kagi: Best paid search option if you would rather pay money than pay with data.
- Startpage: Best for Google-like results without directly searching through Google.
- DuckDuckGo: Best mainstream easy switch for most people.
- Mojeek: Best independent-index backup when you want a different view of the web.
What Google AI Search Changes

Google introduced AI Overviews to generate answers at the top of many search results. In May 2024, Google said AI Overviews would roll out to everyone in the United States and that it expected to bring the feature to more than a billion people by the end of that year.
Google describes this as a way to let Search do more work for users. The company says AI Overviews help people ask more complex questions, get quick summaries, and discover a greater diversity of websites.
That is the friendly framing. The more serious version is this: Google is moving from being a gateway to websites to being the place where more answers are consumed directly.
That matters because search queries are not random. They reveal health fears, legal problems, political curiosity, financial stress, relationship issues, business plans, and private intent. AI search encourages longer, more conversational, more revealing queries. Instead of typing a few keywords, people ask full personal questions.
That makes search privacy more important, not less.
Why People Are Angry About AI Overviews
The frustration is not just that Google added AI. The frustration is that Google placed AI summaries into a product people already depend on, then made the classic list of links feel secondary.
For many users, search is supposed to do one simple thing: show the best places to go next. AI Overviews change the experience into something closer to a platform-controlled summary.
That creates four problems:
- User agency drops. You see Google’s synthesized answer before choosing your own sources.
- Publisher traffic can fall. If the answer is already on Google, fewer people click through.
- Source context gets flattened. A cited source is not the same as a visited source.
- Search privacy gets more sensitive. AI prompts can reveal more than normal keywords.
The Click Problem: Sources Get Cited, But Do They Get Traffic?

Pew Research Center analyzed browsing data from 900 U.S. adults who agreed to share online activity. Pew found that users clicked traditional search results on 8% of Google search visits when an AI summary appeared, compared with 15% of visits when no AI summary appeared. Pew also found that users clicked a source link inside an AI summary only 1% of the time.
That is the key issue. A website can be credited and still lose the visit, the ad impression, the newsletter signup, the donation, the affiliate click, or the reader relationship.
Ahrefs also analyzed 300,000 keywords and reported that AI Overviews correlated with a 34.5% lower average click-through rate for the top-ranking page compared with similar informational keywords without an AI Overview. That is not a randomized experiment, so treat it as directional, not absolute. Still, it lines up with the common publisher fear: AI answer boxes can absorb the value of informational searches.
Google disputes the doom narrative. It says links in AI Overviews can get more clicks than if the page appeared as a traditional listing for that query. The problem is transparency. Publishers still need clean reporting that separates AI Overview impressions, AI Overview citations, AI Overview clicks, and standard organic search data.
What This Means For The Internet

The old web bargain was simple:
- People publish useful pages.
- Search engines index those pages.
- Users click through to the source.
- Publishers earn traffic, trust, revenue, or audience.
AI search risks changing that bargain:
- People publish useful pages.
- Search engines ingest and summarize those pages.
- Users stay on the search engine.
- Publishers get fewer direct relationships with readers.
If that becomes the default, the open web gets weaker. Small sites are hit hardest because they often depend on search discovery. Big brands can survive through apps, subscriptions, direct traffic, newsletters, and legal pressure. Smaller writers, hobby sites, local guides, independent reviewers, forums, and niche experts have fewer escape routes.
That is why this is not just a search feature. It is an internet infrastructure issue.
Why You Should Care Even If You Like AI
You can like AI tools and still reject forced AI layers in search.
The issue is not whether AI summaries are sometimes useful. They are. The issue is whether one company should control the answer layer for the public web while also running the ad market, browser, mobile operating system, analytics stack, video platform, and search monopoly.
When a search engine becomes the answer engine, it gains more power over:
- Which sources are visible.
- Which claims are summarized.
- Which publishers survive.
- Which questions are considered answered.
- How much context users reveal in queries.
Convenience is real. So is centralization.

Best Privacy Search Alternatives

| Option | Best for | Account required? | Main tradeoff |
|---|---|---|---|
| SearXNG | Self-hosted metasearch and control | No | Needs setup and maintenance |
| Brave Search | Easy Google replacement with independent index | No | Cloud service, AI features still provider-run |
| Kagi | Paid search with cleaner incentives | Yes | Costs money |
| Startpage | Google-like results with less direct Google tracking | No | You still trust Startpage |
| DuckDuckGo | Mainstream private search switch | No | Depends heavily on partners such as Bing |
| Mojeek | Independent-index research diversity | No | Smaller index |
| MetaGer | Nonprofit metasearch | No | Quality can vary by region and upstream engines |
SearXNG
SearXNG is a free metasearch engine. It aggregates results from many search services and is designed to avoid tracking or profiling users. It can be used through public instances, but the strongest privacy benefit comes from running your own private instance.
Best for: people who want control, self-hosting, and a search engine they can configure.
Downside: you need a VPS or home server, Docker or another install method, and basic maintenance. Upstream engines may also rate-limit requests.
Brave Search
Brave Search is one of the easiest serious replacements because it has its own independent index and does not require you to self-host anything. It includes AI features, but it is not built on the same Google results dependency as many alternatives.
Best for: users who want a quick default search replacement.
Kagi
Kagi is paid search. That turns some people away, but it also fixes the incentive problem. If a search company is paid by users, it does not need to monetize attention the same way an ad-funded engine does.
Best for: people who would rather pay with money than with behavior data.
Startpage
Startpage is useful if you still want Google-like result quality but do not want to search directly through Google. It is not self-hosted, so you still trust Startpage as a middle layer.
DuckDuckGo
DuckDuckGo is the easiest mainstream privacy switch. It is not perfect, and it depends on partners for much of its search data, but it is still a practical step away from Google tracking for everyday users.
Mojeek and MetaGer
Mojeek matters because it runs its own crawler and index. MetaGer matters because it is a nonprofit metasearch option. Both are useful if you want to escape the Google and Bing monoculture.
How To Run SearXNG On A VPS

The cleanest setup is a small VPS with Docker Compose and a reverse proxy such as nginx or Caddy. Do not expose a personal SearXNG instance carelessly. If it is public, it can be abused. For personal use, put it behind a domain, private network, or access control.
1. Install Docker
sudo apt update
sudo apt install -y ca-certificates curl gnupg docker.io docker-compose-plugin
sudo systemctl enable --now docker
sudo usermod -aG docker "$USER"
Log out and back in so the Docker group change applies.
2. Create the SearXNG folder
mkdir -p ~/searxng/core-config
cd ~/searxng
3. Use the current official container template
curl -fsSLO https://raw.githubusercontent.com/searxng/searxng/master/container/docker-compose.yml
curl -fsSLO https://raw.githubusercontent.com/searxng/searxng/master/container/.env.example
cp -i .env.example .env
Edit the .env file. For a reverse proxy on the same server, binding to localhost is usually safest:
SEARXNG_HOST=127.0.0.1
SEARXNG_PORT=8080
4. Add basic settings
nano ~/searxng/core-config/settings.yml
Use this as a starting point:
use_default_settings: true
general:
debug: false
instance_name: "Private SearXNG"
search:
safe_search: 0
autocomplete: "duckduckgo"
formats:
- html
server:
secret_key: "REPLACE_WITH_LONG_RANDOM_SECRET"
base_url: "https://search.example.com/"
limiter: true
image_proxy: true
public_instance: false
method: "POST"
valkey:
url: valkey://searxng-valkey:6379/0
Generate a real secret:
openssl rand -hex 32
5. Start it
docker compose up -d
docker compose ps
Test locally:
curl http://127.0.0.1:8080/
Reverse Proxy With Nginx

Point a subdomain such as search.example.com to your server, then add an nginx site:
server {
listen 80;
server_name search.example.com;
location / {
proxy_pass http://127.0.0.1:8080;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
}
}
Then add HTTPS:
sudo apt install -y certbot python3-certbot-nginx
sudo certbot --nginx -d search.example.com
Caddy is simpler if you prefer automatic HTTPS:
search.example.com {
reverse_proxy 127.0.0.1:8080
}
Can You Run SearXNG On Android With Termux?

Yes. It is possible, but it is not the path I would recommend for a public search instance. A properly secured VPS is more stable. A phone setup is best for personal experiments, travel, emergency use, or learning how the stack works.
Option A: Native Termux install from Git
This uses the same basic idea as the official non-Docker SearXNG install: clone the repository, create a Python environment, install dependencies, point SearXNG at a settings file, and run the web app locally.
pkg update && pkg upgrade
pkg install -y git python clang rust libxml2 libxslt openssl libffi
mkdir -p ~/apps
cd ~/apps
git clone https://github.com/searxng/searxng.git
cd searxng
python -m venv ~/searxng-venv
. ~/searxng-venv/bin/activate
pip install -U pip setuptools wheel
pip install -U pyyaml msgspec typing-extensions pybind11
pip install --use-pep517 --no-build-isolation -e .
mkdir -p ~/.config/searxng
cp utils/templates/etc/searxng/settings.yml ~/.config/searxng/settings.yml
openssl rand -hex 32
Edit ~/.config/searxng/settings.yml. At minimum, set a long random server.secret_key. For phone use, keep it local unless you know exactly what you are exposing.
export SEARXNG_SETTINGS_PATH="$HOME/.config/searxng/settings.yml"
cd ~/apps/searxng
python -m searx.webapp
Then open:
http://127.0.0.1:8888
Option B: Debian in Termux with proot-distro
If native Termux dependencies fight you, use a Debian userspace. This is heavier, but it behaves more like the official Linux install path.
pkg update
pkg install -y proot-distro
proot-distro install debian
proot-distro login debian
apt update
apt install -y git python3-dev python3-venv python-is-python3 build-essential libxslt-dev zlib1g-dev libffi-dev libssl-dev
useradd --shell /bin/bash --home-dir /usr/local/searxng --create-home searxng
su - searxng
git clone https://github.com/searxng/searxng /usr/local/searxng/searxng-src
python3 -m venv /usr/local/searxng/searx-pyenv
. /usr/local/searxng/searx-pyenv/bin/activate
pip install -U pip setuptools wheel pyyaml msgspec typing-extensions pybind11
cd /usr/local/searxng/searxng-src
pip install --use-pep517 --no-build-isolation -e .
From there, create a settings file and run python -m searx.webapp with SEARXNG_SETTINGS_PATH set, like the native option.
Keeping it alive on Android
Android may kill background work. The most stable workarounds are:
- Disable battery optimization for Termux.
- Install Termux:API and run
termux-wake-lock. - Run SearXNG inside
tmuxso the process survives disconnects. - Keep the instance bound to
127.0.0.1for local phone use. - Use Tailscale or WireGuard if you need private access from other devices.
pkg install -y tmux termux-api
tmux new -s searxng
termux-wake-lock
export SEARXNG_SETTINGS_PATH="$HOME/.config/searxng/settings.yml"
cd ~/apps/searxng
. ~/searxng-venv/bin/activate
python -m searx.webapp
What about Docker on Android?
Docker on Android usually requires root, a custom kernel, or Linux container workarounds. If you already run a rooted Android/Linux environment, you can try it. For most people, native Termux or proot-distro is more realistic. For anything public or daily-use, a $5 VPS with nginx or Caddy is still the sane choice.
How To Make SearXNG Your Browser Search Engine

Firefox Desktop
- Visit your SearXNG instance.
- Open Firefox search settings.
- Add the site as a search engine if Firefox detects it.
- Set it as your default.
Chrome or Brave Desktop
Open:
chrome://settings/searchEngines
Add a site search engine:
Name: SearXNG
Shortcut: sx
URL: https://search.example.com/search?q=%s
Phone
Firefox Android is usually easier than Chrome or Brave for custom search. If your mobile browser fights custom defaults, add your SearXNG instance to your home screen and use it like an app.
Final Recommendation
If you want the easiest privacy upgrade, switch your default search to Brave Search, DuckDuckGo, Startpage, or Kagi.
If you want control, run SearXNG on a VPS and use it from every device.
If you care about the open web, do not let one company turn every website into raw material for its answer box. Search should help you find sources. It should not replace the habit of visiting them.
Follow The Thrifty Dev
If you care about privacy, open tools, and building outside Big Tech’s walled gardens, follow The Thrifty Dev on X, TikTok, and Nostr.
FAQ
Is Google AI Search private?
It is still Google Search, which means your activity can be handled under Google’s privacy policy and account settings. AI search can also encourage more detailed queries, which may reveal more sensitive context.
Can I turn off AI Overviews?
Google does not offer a universal permanent switch that restores classic search for every user and query. Availability and workarounds can change by region, browser, account, and product updates.
What is the best private alternative to Google Search?
For most people, Brave Search, DuckDuckGo, Startpage, or Kagi are easiest. For self-hosting and control, SearXNG is the strongest option.
What is SearXNG?
SearXNG is a free metasearch engine that aggregates results from many search services. It is designed to avoid tracking and profiling users.
Should I use a public SearXNG instance?
Public instances are convenient, but you must trust the instance operator. A private instance gives you more control.
Citations and Further Reading
- Google: Generative AI in Search, May 2024
- Pew Research Center: Google users are less likely to click on links when an AI summary appears
- Ahrefs: AI Overviews Reduce Clicks by 34.5%
- SearXNG documentation
- Startpage privacy policy
- DuckDuckGo privacy policy
- Brave Search
- Kagi
- Mojeek
- MetaGer
Private Search Alternatives Comparison
| Search engine | Features | Privacy | Self-hosting |
|---|---|---|---|
| Best index, AI Overviews | High tracking by default | No | |
| DuckDuckGo | Simple private search | Better privacy, Microsoft dependency | No |
| Brave Search | Independent index, Goggles | Strong privacy posture | No |
| SearXNG | Metasearch, customizable | Best when self-hosted | Yes |
Self-Host SearXNG in 10 Minutes
docker run -d --name searxng -p 8080:8080 searxng/searxng:latest
Get the Private Search Setup Guide
Get the practical escape map: search, email, cloud files, photos, browser, phone backups, and what to replace first.
No spam. Practical privacy, AI, backup, and tool drops. Unsubscribe anytime.Related Posts
- Run Your Own AI: The Beginner’s Guide to Local LLMs in 2026
- Private AI Is Becoming the New Developer Edge: Why Venice AI Fits the 2026 Shift
- n8n AI Agents: Self-Hosted Automation Guide (2026)
Keep building from here
These related guides continue the same thread with practical next steps:
- Privacy and Digital Rights Hub
- Private AI Is Becoming the New Developer Edge: Why Venice AI Fits the 2026 Shift
- Run Your Own AI: The Beginner’s Guide to Local LLMs in 2026
- Mandatory ID Is Coming for Phones and Social Media. Here’s How to Move to Nostr Before the Gate Closes
- Meshtastic: Build Your Own Off-Grid Mesh Network
- FCC Wants Your ID to Activate a Phone — Privacy Is Dead
Views: 5