Artificial intelligence is becoming part of everyday life. People are using AI to research coins, jewelry, sterling silver, antiques, collectibles, and just about everything else that might be sitting in a drawer, safe-deposit box, basement, or inherited collection.

We use AI too.

It is an extraordinary tool, and we are optimistic about where it is going. But right now there is an important problem that doesn’t get discussed enough:

AI inherited the Internet we built.

And for roughly the last 25 years, a significant portion of that Internet was created not simply to provide the best possible information, but to attract clicks.

That distinction matters enormously when you’re asking an AI what something is worth.

The Internet Wasn’t Built as a Giant Encyclopedia

It’s easy to look at today’s Internet and imagine that billions of webpages gradually accumulated because people had useful information they wanted to share.

Some certainly did.

But anyone who worked with websites during the earlier days of search engines remembers that there was another game being played.

At one point, search engine optimization could be remarkably primitive. If you wanted a page to appear when somebody searched for a particular phrase, one strategy was essentially to put that phrase on the page as many times as you could without making the page completely unreadable.

If you wanted to rank for “Chicago coin dealer,” you might write something resembling:

“If you’re looking for a Chicago coin dealer, our Chicago coin dealer can help you sell coins to a Chicago coin dealer serving customers looking for the best Chicago coin dealer…”

It sounds ridiculous today.

But variations of this actually worked.

Website owners experimented with repetitive keywords, exact-match domains, doorway pages, link exchanges, oversized networks of related pages, and all sorts of other techniques designed primarily to convince search engines that one webpage deserved to appear above another.

Search engines improved. Website owners adapted. Search engines improved again. Publishers adapted again.

That cycle continued for decades.

Google Rewarded Some Strange Behavior Over the Years

This isn’t really a criticism of Google. Search engines were attempting to solve an extraordinarily difficult problem: organizing an exploding Internet and figuring out which pages were actually useful.

But every ranking system creates incentives.

When Google rewarded links, people built links.

When longer articles performed well, people wrote longer articles.

When certain questions attracted traffic, publishers created hundreds or thousands of pages answering those questions.

When dramatic headlines attracted clicks, publishers discovered dramatic headlines.

And collectibles were almost perfectly designed for that environment.

“This Penny Could Be Worth $1 Million!”

Consider two possible headlines:

“Most Old Pennies You Find Are Common and Worth Very Little“

or:

“CHECK YOUR POCKET CHANGE: This Penny Could Be Worth $1 Million!”

Which one would you click?

Publishers figured that out a long time ago.

There really are extraordinary coins. There really are rare varieties. There really are coins worth tens of thousands, hundreds of thousands, or even millions of dollars.

That’s what makes the headline technically possible.

The problem is that the extremely valuable example may have almost nothing to do with the ordinary coin sitting on someone’s kitchen table.

The same thing happens with old paper money, sterling silver, jewelry, toys, medals, watches, antiques, sports cards, and countless other collectibles.

The exceptional item gets the article.

The ordinary item usually doesn’t.

Twenty Websites Don’t Necessarily Mean Twenty Sources

This may be one of the most important things to understand about information on the Internet.

Imagine that you search for a collectible and find 20 websites making essentially the same claim about its value.

That sounds convincing.

Twenty different sources agree.

Except they might not actually be 20 independent sources.

One article may have referenced an old price guide. Five other websites may have rewritten that article. Ten more may have rewritten those five. Another site may have summarized everything for search traffic.

Suddenly one original claim has become 20 webpages.

Nothing necessarily dishonest happened. The information simply reproduced itself.

This is one of the enormous challenges facing modern AI.

Finding information is no longer the difficult part. Figuring out where that information originated, how independent it is, and how much weight it deserves is much harder.

AI Has a Lot of Information. That Doesn’t Mean It Has Completely Put It Together Yet.

Modern AI systems can work with an extraordinary amount of human knowledge. They can compare information, summarize complicated subjects, recognize patterns, explain terminology, and find connections that would take a person hours or days to research manually.

That’s remarkable.

But AI is still relatively new.

Having enormous amounts of information available is different from perfectly understanding how every piece fits together.

Collectibles demonstrate this problem particularly well.

A computer might correctly determine that a particular coin exists and that an example sold for $25,000.

But that doesn’t necessarily mean the coin in your hand is worth $25,000.

The expensive example might have been a rare variety. It might have been in extraordinary condition. It might have had an important provenance. It might have been certified by a major grading service. It might have been an unusual error.

Or your coin might simply look similar.

Those distinctions can represent the difference between a few dollars and thousands of dollars.

Rarity Doesn’t Automatically Mean Value

Another concept that can be difficult for both humans and AI is the relationship between rarity and demand.

People naturally assume that if something is extremely rare, it must be extremely valuable.

Sometimes that’s true.

A famous low-mintage coin with thousands of collectors competing for a small number of surviving examples can be extraordinarily valuable.

But there is a strange flip side to rarity.

Some objects are so obscure that almost nobody knows they exist.

Something can be genuinely rare and still have a very small market because very few people collect it.

Coin and collectibles dealers encounter this more often than most people would expect.

Rarity is only one part of value.

You also need demand.

Asking Price Is Not Market Value

This is another enormous source of confusion online.

Someone can ask almost anything for an item.

A coin listed online for $5,000 is not evidence that somebody paid $5,000 for it.

A collectible sitting unsold at $1,500 for three years isn’t necessarily a $1,500 collectible.

Actual transactions generally tell us much more than asking prices.

Even completed sales need interpretation. Condition, certification, authenticity, date, variety, provenance, buyer premiums, market timing, and other factors can dramatically change the meaning of a comparable sale.

This is why valuation is much more complicated than simply finding a similar-looking object online.

Retail Value, Auction Value, Wholesale Value and Melt Value Aren’t the Same Thing

Another problem is that the word “value” can mean several completely different things.

An insurance replacement value may be one number.

A retail asking price may be another.

An auction result may be another.

A dealer’s wholesale buying price may be another.

The value of the precious metal contained in an item may be another. That figure is generally based on the current spot price of gold or silver.

None of these numbers is automatically wrong.

They’re answering different questions.

If someone asks an AI, “What is this worth?” without providing enough context, the AI has to determine which question the person actually means.

Humans have exactly the same problem.

The Internet Has 25 Years of Collectibles Clickbait

This is where things get particularly interesting.

AI didn’t invent sensational stories about valuable pennies.

AI didn’t invent exaggerated antique values.

AI didn’t invent endless articles about supposedly priceless items hiding in your attic. We did…..

For decades, publishers discovered that stories about unexpectedly valuable objects attracted readers.

Search engines sent traffic to those stories.

Advertising turned traffic into money.

Other publishers saw the traffic and created similar stories.

Social media amplified the most surprising examples even further.

Video platforms added another layer.

And now AI has arrived and been handed the entire pile.

AI’s Next Challenge Is Understanding the Genealogy of Information

We think this is where AI could eventually become much better than traditional Internet search.

The goal shouldn’t simply be counting how many webpages repeat a claim.

A truly useful system needs to understand the genealogy of information.

Where did the claim originate?

Are ten articles actually ten independent sources, or are nine of them repeating the first one?

Is a price based on an actual completed transaction?

Was it an asking price?

Was the information written by someone with firsthand experience?

Is the source discussing retail prices, wholesale prices, auction prices, insurance values, or precious-metal value?

Is the information current?

Does an actual marketplace support the claim?

Those are much harder questions than simply finding a webpage containing the right words.

There’s an Important Difference Between Searching Google and Asking AI

For many years, the basic relationship between a person and a search engine was fairly obvious.

You asked Google a question.

Google gave you a list of websites.

You clicked those websites and decided which information you trusted.

That relationship is changing.

Today someone can ask a question and receive an AI-generated explanation directly within the search experience.

To many users, that simply feels like Google Search got better.

And in many ways it did.

But something important has changed.

The computer is no longer simply helping you find information.

It is increasingly helping you interpret information.

That is an enormously useful development, but it also makes understanding sources and uncertainty more important.

Don’t Ignore the Disclaimer

AI companies themselves warn users that AI can make mistakes.

That’s not fine print that should simply be ignored.

Our approach is straightforward:

Assume that anything important an AI tells you may need to be verified.

That doesn’t make AI useless.

Quite the opposite.

A calculator is useful even though you can enter the wrong numbers. A search engine is useful even though bad websites exist. A price guide is useful even though markets change.

AI is another tool.

It happens to be an extraordinarily powerful one.

But a powerful research tool isn’t automatically the final authority.

We Actually Think AI Can Make This Better

Despite everything we’ve just written, we’re optimistic.

The old Internet created an enormous information-quality problem.

AI may eventually help solve it.

Instead of forcing someone to open 30 search results, compare contradictory articles, determine which websites copied each other, interpret auction records, understand specialized terminology, and somehow reach a conclusion, AI can potentially do much of that work.

But doing it well requires more than collecting information.

It requires understanding which information deserves the most weight.

That capability will continue to improve.

Why We Put Firsthand Information on This Website

Before working in coins and collectibles, I worked in web and graphic design. I’ve had the unusual experience of watching the Internet evolve from both sides: first as someone building websites and later as a coin and collectibles dealer watching customers use those websites to research their possessions.

Now we’re watching another transformation as AI becomes part of that process.

That’s one reason we try to publish information based on what we actually encounter at Oakton Coins & Collectibles.

When we discuss inherited coin collections, precious metals, sterling silver, jewelry, bullion, paper money, or unusual collectibles, we’re not simply rewriting another article because a search engine says people are looking for that phrase.

We’re trying to document what happens in the real marketplace.

What do people actually bring through the door?

What actually sells?

What doesn’t?

What mistakes do sellers repeatedly make?

What looks rare but isn’t?

What really is rare but has surprisingly little demand?

What happens when something has to be sold wholesale?

What happens when precious metals have to be refined?

Those experiences are another form of data.

And as AI becomes increasingly responsible for organizing the world’s information, we think firsthand information will become more important, not less.