Originally published: September 5, 2026 on Truth Decay by James Macleod.
Let’s start with an uncomfortable fact dressed as a magic trick.
If you sit two people down at two identical laptops, in the same room, at the same second. Have them type the exact same six words into the exact same search engine.
They will not get the same results back.
Not similar. Not roughly comparable, give or take an advert. Completely. Different.
Because the version of the web that gets shown to you isn’t really the web. It’s a personalised, monetised, algorithmically bartered guess about what you, specifically, can be persuaded to click on, and ideally, buy.
Google isn’t simply reading your search query. It is reading you, then cross-referencing the query against the many gigabytes of data it already has about you, and then quietly decides which slices of human knowledge is commercially worth showing you today.
Which is both a remarkable achievement when you think about it, and absolutely terrifying when you think about it just a little bit more deeply.
We invented the world’s largest information network and then handed the front door to an advertising company.
And then we acted surprised when there was a gift shop.
This is not a conspiracy theory. It’s the business model.
Search engines make money when you click on things that make somebody money. A beautifully written, meticulously researched personal blog with zero advertising has, from a search engine’s perspective, roughly the commercial appeal of a discontinued product.
Nobody’s bidding on those keywords.
Nobody’s paying for that placement.
So down the rankings it goes. Past page one. Past page ten. Into the sort of internet Siberia where only the truly stubborn, the professionally curious and people who have forgotten what sleep is ever venture.
The web, in other words, is enormous, and what you’re shown is a tiny, curated, for-profit sliver of it.
The good news is that the rest of it still exists.
It hasn’t been deleted.
Mostly.
We’ll get to that.
It’s just been buried under several tectonic layers of search engine optimisation, commercial incentives, algorithmic prioritisation, advertising and websites that appear to have been designed by a committee whose only instruction was “please make us money”.
Somebody has to actually go looking for the shovel.
So, here’s the shovel.
Marginalia: The Anti-Algorithm.
If Google is a shopping centre with a very loud and insistent PA system, Marginalia Search is the second-hand bookshop three streets over that doesn’t have a sign, doesn’t want a sign, and would honestly prefer you didn’t tell your friends about it.
Marginalia is built and run, largely single-handedly, by a Swedish developer who got pissed off watching the same handful of SEO-optimised, ad-choked, content-farmed websites dominate every search result for every conceivable query.
So he did something almost offensively unfashionable, or in my eyes, truly nerd gangster genius…
He built a search engine that tries to do the opposite.
Marginalia runs its own independent crawler and index rather than piggybacking on Google or Bing’s infrastructure, and it deliberately favours text-heavy, non-commercial, personally handcrafted websites over the sleek, JavaScript-laden, cookie-banner-festooned professional web most of us have been trained to tolerate.¹
It is, delightfully, a search engine that punishes good web design.
This is perhaps its greatest selling point.
The internet is full of websites so polished that you can practically hear a venture capitalist breathing behind the homepage. Marginalia has other priorities.
It rewards the sort of site your nerd uncle Ned built in 2004 and never touched again, on the theory that somebody who built something with genuine care and no obvious commercial motive might actually have written something worth reading.
Imagine that.
A website made by a human because the human had something to say. Revolutionary stuff.
The project survives on European public-interest grants rather than advertising revenue, which means nobody is trying to sell you a mattress via your search results.²
Try searching for a niche hobby, an obscure historical event, or literally anything you would expect Pinterest and Reddit to smother in the regular results.
What comes back looks like the internet used to look.
Personal.
Weird.
Occasionally badly formatted.
Entirely uninterested in converting you into a customer.
There may be no beautifully optimised landing page. No “10 Things You NEED To Know About Medieval Cheese”. No lifestyle influencer standing beside a bowl of something beige.
Just information.
Actual information.
Practical tip: Marginalia rewards specific, unusual phrasing over broad keywords.
Vague queries get vague results everywhere. Here, oddly specific questions are where it shines.
It’s less a search engine than a very well-read stranger at a party who has actually read the book instead of the review.
And, crucially, the stranger isn’t trying to sell you a mattress.
Find it at search.marginalia.nu.
Million Short: Subtraction As A Search Strategy.
If Marginalia rebuilds the web from scratch to favour the obscure, Million Short takes a considerably blunter approach.
It lets you physically remove the most popular results from view and see what’s left standing.
The tool lets a user strip out the top 100, 1,000, 10,000, 100,000, or full million most prominent sites from any given search, on the reasoning that if you’ve genuinely never seen anything beyond the first page of Google in your life, you might as well find out what’s hiding on page four thousand.³
It is, essentially, search engine gardening.
Pull out the weeds and see what grows.
Search your own name with the top million results deleted and you’ll likely be quietly relieved, mildly horrified, or both, to discover what surfaces once Facebook, LinkedIn and the rest of the usual suspects are forcibly evicted from the results.
It’s less a serious research tool than a philosophical statement rendered in code.
An entire functioning search engine built on the premise that “popular” and “relevant” are not the same thing.
That is a surprisingly radical proposition for the modern internet.
We have somehow reached a point where the largest websites have occupied the top spots for so long that “the internet” and “the websites Google thinks are important” have become almost interchangeable concepts.
They aren’t, and this should really matter to us more than is seems to.
Because popularity is not truth.
Popularity is not expertise.
Popularity is not even necessarily usefulness.
Sometimes it is simply what has the biggest marketing department.
Check it out:
BASE: Where The Actual Research Lives.
Speaking of things squatting on the top spots, here’s a fun experiment.
Search any academic topic on Google and watch how quickly you hit a paywall.
Somewhere behind that paywall sits a peer-reviewed study that took years to produce, probably involved several researchers and very likely consumed public grant money, and could potentially tell you something genuinely important about the world.
You are now being asked for £35 to read the abstract.
Science has never been more accessible.
Please enter your credit card details.
This is where BASE, the Bielefeld Academic Search Engine, quietly does what Google won’t: it helps you find the often publicly funded research, without charging admission.
Run out of Bielefeld University Library in Germany, BASE indexes well over a hundred million academic documents pulled from thousands of institutional repositories worldwide, and roughly sixty percent of what it indexes is available in full text, for free, entirely legally.⁴
No login wall pretending to be a paywall pretending to be a “free trial”.
No countdown clock informing you that your “limited access” expires in 17 minutes.
Just the paper.
It’s not glamorous.
It looks like it was designed in 2009, because large parts of it were.
And frankly, good.
There is something comforting about a website that looks as though nobody has recently held a meeting called “How Do We Monetise The User Journey?”
If you’ve ever wanted to read the actual underlying study behind a headline instead of a journalist’s three-paragraph gloss of a press release, BASE is where you go to find out whether the science actually said what everyone claims it said.
Because oftentimes the headline says one thing, the press release says another, the article says something else entirely, and the actual paper is sitting sheepishly in the corner wondering how the hell it even got dragged into this.
https://www.base-search.net/
The Vanishing Internet, And The One Organisation Trying To Stop It.
Here’s where things get less whimsical.
Because it isn’t only that useful information gets buried by algorithms uninterested in it commercially.
Increasingly, information simply just… Disappears.
Articles get quietly edited after publication with no changelog.
Inconvenient stories get deleted outright.
Corporate and government pages that once said one thing now say another, or say nothing at all.
And there’s no public record that the original version ever existed.
Unless somebody happened to save it first.
That’s precisely the gap the Internet Archive’s Wayback Machine exists to fill.
It is a free, publicly accessible, running snapshot of the web, currently holding well over a trillion archived pages, allowing anyone to check what a website actually said at a given moment in time, regardless of what it has been quietly rewritten to say since.⁵
It is, without exaggeration, one of the single most important accountability tools that exists for journalists, researchers, historians, investigators and anyone who has ever been told:
“That’s not what it said.”
And needed proof that, actually, yes, it bloody well was.
The internet has given humanity an extraordinary ability to record almost everything.
It has also given gatekeepers an extraordinary ability to quietly rewrite almost anything.
The Wayback Machine is one of the few places where the second part can be challenged by the first.
And it is, right now, under genuine strain.
The Internet Archive doesn’t run on government funding the way you might assume a project of this scale would.
It’s a nonprofit that survives almost entirely on individual donations, philanthropic grants and support from foundations like the Kahle-Austin Foundation, operating on an annual budget that is startlingly modest for an organisation holding hundreds of petabytes of the world’s digital memory.⁶
Meanwhile, the wider ecosystem it depends on, including the American library and archive funding infrastructure that supports projects like it, has spent the past two years under sustained political assault.
The federal agency that channels the bulk of US government support to libraries and archives, the Institute of Museum and Library Services, has faced repeated attempts at elimination via executive order, emergency court injunctions, restored funding and then fresh budget proposals threatening to gut it again to a symbolic fraction of its prior size.⁷
Meanwhile, the Archive spent the better part of four years and an undisclosed but reportedly substantial settlement sum fighting a copyright lawsuit brought by major publishers over its digital lending programme.
It ultimately lost at the Second Circuit and chose not to escalate the case to the Supreme Court.⁸
None of this means the Wayback Machine is about to vanish tomorrow.
It means something rather more uncomfortable.
An organisation doing genuinely irreplaceable public-interest work is operating in a funding and legal environment that has grown noticeably more hostile, while quietly asking individual users for donations to keep the lights on.
Think about the absurdity of that for a moment.
We have built an information system capable of storing the accumulated digital output of modern civilisation, and one of the organisations doing the actual work of remembering it is essentially standing outside the internet with a collection bucket.
If you have ever used the Wayback Machine, and if you read investigative journalism, the odds are extremely high that you have benefited from it without realising. If you regularly read Truth Decay, you absolutely have, so it might be worth being one of those donors.
Because once the original version of something disappears, proving that it ever existed becomes considerably harder.
And history has always been rather fond of inconvenient evidence.
Guardrails, Some Useful, Some Deeply Unsettling: A Note On Faces.
This next section is not a recommendation.
I want to be very clear about that up front, because what follows sits somewhere between a warning and a horror story, and I’d rather you read it as the latter.
Companies like Google unquestionably possess the technical capability to make any face on the internet searchable by name, at scale, right now.
They don’t offer this publicly, and there are good reasons to assume that’s a deliberate, considered choice rather than a technical limitation.
They absolutely can do it for their own use.
And that’s still creepy A F.⁹
Whatever Google does with that capability internally, or however it might feed into advertising profiles sold onward to other companies, happens well out of public view.
That is the uncomfortable part about modern surveillance technology.
The important question isn’t always what a company allows you to do.
It is what the company itself is technically capable of doing.
What is public, and considerably less restrained, is a small industry of facial search tools with far fewer guardrails.
If you upload a photograph to a service like PimEyes, it will crawl the public web for matching faces, returning links to every site where that face has appeared, from professional headshots to, more unsettlingly, photographs the person themselves may never have known existed.¹⁰
Privacy regulators in multiple countries have formally complained about it.
Journalists have used it.
So, have stalkers.
Which is not exactly the sort of three-part endorsement any tech company puts on its homepage.
I’ll admit, with some discomfort, to having searched my own face on it out of morbid curiosity more than once, mostly to see what a determined stranger could find.
And I was genuinely unsettled by how much came back.
TinEye sits at the tamer end of the same spectrum, generally used for tracking down where a specific image originated or has been reused, without the same facial-matching ambitions.¹¹
Then there’s Yandex, Russia’s dominant search engine, which has never exactly behaved as though privacy was going to be its defining lifestyle choice.
Yandex’s image search doesn’t just find visually similar pictures the way Google’s does.
It identifies specific faces.
And according to investigative researchers who have stress-tested it extensively, it does so with genuinely alarming accuracy, particularly for anyone with a digital footprint in Europe or the former Soviet sphere.¹²
For what it’s worth, my own casual poking around hasn’t produced results anywhere near that frightening.
But I’d chalk that up to limited testing rather than any actual reassurance.
Professionals who use these tools for a living rate Yandex as the most capable facial search engine currently operating in public.
Your mileage, and apparently your face’s discoverability, may vary.
The broader point isn’t:
“Go and use these.”
It is that the guardrails currently keeping any of this in check are wildly inconsistent, largely self-imposed by the companies involved, and vary just as wildly by jurisdiction.
That should probably concern you more than it currently does.
Because the technology is no longer science fiction.
The question is who gets to use it, under what rules, with what oversight, and how much of the resulting capability is visible to the people being searched.
The answer, at present, is not exactly filling me with the warm glow of democratic accountability.
Shodan: Google, But For “Things” Instead Of Pages.
One more.
And this one’s really quite strange the first time you encounter it.
Every search engine discussed so far indexes web pages.
Shodan indexes devices.
Specifically, it continuously scans the internet for devices that are directly connected to it.
And there are far more of these than you’d think.
Security cameras.
Industrial control systems.
Home routers.
Smart fridges.
And occasionally things considerably more alarming than smart fridges.
Shodan then catalogues what it finds by location, software version and open ports.¹³
It was built for cybersecurity researchers who need to find and patch vulnerable systems before someone with worse intentions does.
That’s a genuinely useful, arguably essential function.
It is also, depending entirely on who’s doing the searching and why, an extraordinarily effective way to locate exactly which unsecured webcams and industrial systems in a given city or country are sitting there, unprotected, waiting to be found.
The same tool that helps a security researcher patch a vulnerability helps someone with fewer scruples exploit it first.
Shodan doesn’t have an opinion about which of you shows up.
It is, in that sense, a perfect example of the internet itself.
The same technology can be immensely useful and deeply unsettling depending on who has access to it and what they intend to do with it.
The internet has never been particularly interested in moral philosophy.
It has mostly been interested in whether the port is open.
DuckDuckGo: The Duck Has A Microsoft Problem.
At this point you might reasonably be thinking: fine, I’ll just use DuckDuckGo.
And that’s actually not a bad idea.
DuckDuckGo has spent years positioning itself as the privacy-conscious alternative to Google, and there is a substantial difference between its stated privacy model and Google’s. DuckDuckGo says it doesn’t save or share your search or browsing history, doesn’t build personal profiles from your searches, and doesn’t personalise search results using personal information.¹⁴
All of which sounds rather lovely.
But there is a small complication.
The duck isn’t entirely flying solo.
Most of DuckDuckGo’s traditional search results come from Microsoft’s Bing. DuckDuckGo maintains its own crawler, DuckDuckBot, and operates various indexes and search features itself, but the conventional blue links that make up most of what you actually see are largely sourced from Bing.¹⁵
So if you’ve just escaped Google because you don’t particularly fancy having your search experience mediated by one enormous American technology corporation, congratulations.
You’ve moved next door.
The house has a different sign on it, the curtains are nicer and nobody is apparently keeping a personalised diary of your search history, but Microsoft still knows where the road is.
That doesn’t make DuckDuckGo pointless.
Quite the opposite.
It makes it interesting.
DuckDuckGo’s model is substantially more privacy-conscious than Google’s, and its search engine doesn’t personalise results around a profile of your previous behaviour in the same way. It is also an independent company rather than a Google subsidiary, and its stated policy is that it doesn’t retain search histories linked to individual users.¹⁶
But “more private than Google” and “independently verifiable privacy fortress” are not the same thing.
DuckDuckGo is still a US company, headquartered in Pennsylvania, and therefore operates within the American legal and regulatory environment.¹⁷ That doesn’t mean the company is secretly handing everything to the U.S. government. There is no evidence presented here that would justify making such a claim.
It does mean that jurisdiction matters.
If a company holds information about you, the laws governing that company, the legal demands it can receive and the circumstances under which it may be required to disclose information are all relevant questions.
And then there is the question of trust.
DuckDuckGo makes the source code for many of its products and privacy technologies available under open-source licences. Its Tracker Radar, tracker blocklists and various browser components can be inspected publicly.¹⁸
But the existence of open-source components doesn’t make the entire DuckDuckGo search operation open source.
You cannot simply download the complete search engine, inspect every line of code, reproduce its entire infrastructure and independently verify every privacy claim yourself.
At some point, you still have to take the company at its word.
And that’s really worth remembering because privacy companies are still companies.
They have employees.
They have contracts.
They have commercial partners.
They have lawyers.
They have shareholders or investors.
And, occasionally, they discover that a business arrangement they signed yesterday is rather less philosophically pure than the privacy manifesto they published the day before.
DuckDuckGo provided a particularly useful demonstration of this in 2022, when a security researcher discovered that its browser allowed certain Microsoft advertising trackers to operate under an exception connected to DuckDuckGo’s search syndication agreement with Microsoft.
DuckDuckGo’s CEO acknowledged the issue, and the company subsequently changed its protections. The incident didn’t demonstrate that DuckDuckGo Search was secretly recording users’ searches in the way Google does. What it demonstrated was something probably more useful:
Even privacy companies can have privacy compromises buried inside the plumbing.¹⁹
That should not be interpreted as “DuckDuckGo is secretly Google.”
It isn’t.
Nor should it be interpreted as “never use DuckDuckGo.”
If anything, it is a great argument for using it.
Just don’t turn a privacy policy into a religion.
A company saying, “trust us, we don’t collect that information” is useful.
A system that makes it technically difficult or impossible for the company to collect that information is better.
And an independently verifiable system is better still.
DuckDuckGo has taken steps in that direction. For example, it commissioned an independent security assessment of its VPN infrastructure covering 2025 and 2026. The assessment included source-code review of proprietary components and live-system analysis and concluded that the VPN infrastructure complied with its stated no-logs policy.²⁰
That’s encouraging.
But it is also important to notice what the audit actually covered.
The VPN.
Not every component of DuckDuckGo Search.
Not every part of the company’s wider ecosystem.
And not every future business arrangement it might enter into.
Which brings us back to the broader point of this article.
The objective isn’t to find a corporation you can trust absolutely.
That’s how we ended up here in the first place.
The objective is to reduce the amount of unnecessary trust you’re required to place in any single corporation.
Use DuckDuckGo if you like.
It’s a reasonable alternative to Google.
Just remember that even the privacy duck has a corporate supply chain.
And somewhere in that supply chain, there’s a Microsoft server quietly quacking into a spreadsheet.
The internet is complicated like that.
Which is why the best defence isn’t finding one magical search engine that promises to protect you from everything.
It’s learning that you have choices.
And, preferably, using more than one of them.
The Bit Where I Tell You What To Actually Do With All This.
Here is where I’m going to depart from the usual polite technology-column ending.
You know the one:
“Of course, we’re not suggesting you stop using Google entirely. It remains an incredibly useful service, and these concerns should simply encourage us to think more carefully about how we use technology.”
No. FUCK. RIGHT. OFF!
Dump Google.
At least as your default.
Yes, Google is extraordinarily good at what it does.
Google Search is fast.
Google Maps is brilliant.
Gmail is useful.
YouTube is effectively the world’s largest video library.
Android is everywhere.
Google has produced an enormous number of genuinely excellently useable products and services that have become deeply embedded in everyday life.
That is precisely why this matters.
The problem isn’t that Google makes bad products.
The problem is that Google makes very good products while operating an extraordinarily powerful data-collection and “advertising” ecosystem.
The convenience is real. So is the cost.
And that cost is not necessarily measured in dollars.
It is measured in data.
Your searches.
Your movements.
Your interests.
Your habits.
Your browsing behaviour.
Your devices.
Your interactions.
Your digital relationships.
Your location history.
Your photographs.
Your videos.
Your emails.
Your preferences.
The things you click.
The things you almost click.
The things you search for at 2am and would rather not explain in a court of law.
Google has moved a very long way from being the slightly nerdy internet company that gave us a better search engine, a quirky collection of useful apps and an alarming number of things named after food.
It is now an enormous technology corporation embedded in some of the most consequential infrastructure on Earth.
And increasingly, that includes the military-industrial complex.
That matters.
Not because every Google product has suddenly become a weapon.
It matters because the scale of information held by companies like Google changes the nature of the power they possess.
When a company knows an extraordinary amount about billions of people, the question is not simply what that company intends to do with the information today.
It is what happens when the company changes.
What happens when executives change?
When ownership changes?
When governments change?
When laws change?
When geopolitical circumstances change?
When technology changes?
When the definition of what is considered acceptable surveillance changes?
Data collected for one purpose has a nasty habit of becoming useful for another.
And information that seems harmless in isolation can become considerably more revealing when enough pieces are assembled.
That is the fundamental problem with handing one corporation an enormous portion of your digital life.
You are not merely trusting today’s Google. And I say that with a very clear memory that todays Google is already the google that dropped the mantra: “Don’t be evil.”
You are trusting every future version of Google that might inherit the machinery.
And that is an awfully large amount of trust to place in a company whose primary business model is built around knowing things about you.
So yes, use the alternatives.
Use Marginalia when you want to find the weird, independent corners of the web.
Use Million Short when you want to kick the giants out of the room and see what survives.
Use BASE when you want the actual research rather than the internet’s press-release interpretation of it.
Use the Wayback Machine when somebody tells you that something never existed.
Be aware that facial-search tools exist and that the technology is considerably more capable than most people realise.
And understand what Shodan demonstrates about the enormous amount of infrastructure that is sitting openly on the internet.
But more broadly:
Stop treating Google and mainstream search as the internet.
It isn’t.
Google is a company.
A very large company.
A very powerful company.
A company with commercial interests, enormous technical capabilities, vast quantities of personal data and increasingly significant relationships with governments and the military-industrial complex.
The internet itself is vastly bigger, stranger, messier and more interesting than the little window Google provides.
The actual internet, the messy, uncommercial, occasionally frightening, occasionally wonderful one, is still there.
It’s indexed by people who built search engines out of stubbornness rather than venture capital.
It’s preserved by an archive fighting for its funding while doing work no government agency has stepped up to replace.
It’s populated by independent writers, researchers, obsessives, hobbyists, academics, archivists and many a lunatic who has spent fourteen years documenting the history of municipal drainage systems.
And sometimes it contains something useful.
You just have to look.
The internet has a basement.
Google has simply spent years convincing us that the basement doesn’t exist.
Open the door.
nice work thanks. keep up the great work.
If you now add a server to the web Google doesnt even find it as GPT is only boxed in its LLM.
And that could be ok if other crawlers still manage to find it, but do they?