AI in the wild: Confident, wrong, and weirdly expensive

Jun 04, 2026 08:00 PM - 1 month ago 47888
AI successful the wild- Confident, wrong, and weirdly expensive-featured-image

Imagine moving successful SEO for years, researching a problem you cognize wrong and out, only to person an LLM confidently explicate why your acquisition is wrong.

That happened to me. Actually, it happened 3 different times past week pinch Gemini.

The scary portion wasn’t that the answers were evidently bad. It was that they sounded good. The responses were polished, believable, and directionally meticulous capable that astir group would ne'er mobility them.

And if you aren’t profoundly acquainted pinch the topic, you astir apt wouldn’t cognize really to situation the answer.

Funny enough, I caught it twice. The 3rd clip costs maine money. All wrong the aforesaid week.

View embedded content

Example 1: Gemini educates maine connected method SEO

Let maine springiness you immoderate context.

I’m presently helping migrate a client’s FAQ hub from a provider-hosted subdomain to a self-hosted implementation.

The FAQ lives nether a /faq/ folder, but individual Q&A pages are parameter-based URLs. Normally not an issue… isolated from Shopify forces canonicals backmost to the guidelines /faq/ page, efficaciously preventing those pages from being indexed.

While researching Shopify-specific solutions and plagiarism considerations, Gemini gave maine this response:

signals

Excuse me?!

You will perfectly not beryllium penalized for conflicting SEO signals.

At best, Google indexes what it wants. At worst, it ignores directives it doesn’t trust.

The bigger rumor is the wording.

“Penalty” is fundamentally the magic connection successful SEO. The infinitesimal activity hears it, priorities shift, momentum dies, and recommendations go harder to implement.

Then Gemini doubled down.

params

I asked whether removing the canonicals and allowing the parameter pages to beryllium independently was an option.

Gemini responded:

  • “Google mostly ignores query parameters.”

Wow… conscionable wow.

Below is an implementation I worked connected pinch the Saatva team, wherever we intentionally indexed parameter URLs wrong the shopping experience.

saatva

Search Console and URL inspection confirmed those pages indexed conscionable fine.

Parameter pages tin perfectly rank, index, and make value.

The problem isn’t that Gemini was incorrect (as successful it’s a much difficult implementation).

It’s that the reply sounded believable capable that personification without SEO acquisition would astir apt judge it and move on.

param-good

Example 2: Gemini says lick the rumor pinch a $3,000 part

This 1 deed otherwise because I’m not a mechanic.

I’ve been troubleshooting an rumor connected my Jeep SRT and spent hours extracurricular successful the sun collecting data, testing fuses, reviewing OBD2 logs, and trying to constrictive down the guidelines cause. After reviewing everything, Gemini confidently concluded the rumor was apt a rear differential nonaccomplishment and suggested a afloat replacement.

Estimated damage? Roughly $3,000 successful OEM parts alone.

rear diff cost

The consequence sounded fantastic. It was detailed, logical, and moreover complimented my troubleshooting process.

The problem? It was wrong. Really wrong.

After pushing backmost and sharing further OBD2 information I had been search during diagnosis, Gemini wholly reversed people and admitted it jumped to a worst-case script without capable evidence.

bad_gemini

In different words, I almost spent thousands replacing parts I didn’t need.

That’s what makes these examples interesting. In SEO, I instantly knew Gemini was incorrect because I had years of acquisition to situation the recommendation. Here, I didn’t person that advantage. I had to trust connected skepticism, proceed testing, and debar treating the reply arsenic fact.

Same AI. Same confidence. Completely different outcome.

Example 3: Gemini costs maine $20 cardinal dollars!

OK, technically, this was video crippled money.

I was playing Madden, moving done squad logistics, and trying to optimize net headdress spending to re-sign cardinal players and support the roster together.

At immoderate point, I sewage lazy and tagged successful my partner, Gemini. I shared a screenshot of the squad finances and asked for a roadmap to restructure contracts and optimize the headdress situation.

Gemini gave maine a elaborate plan. I followed it.

And it put maine $20 cardinal complete the net cap.

The funny portion is this perfectly mirrors the earlier examples.

The reply looked good. It was organized, confident, and gave maine a clear player-by-player action plan.

So I trusted it. Then I called Gemini out.

madden dollars

You’ll announcement successful the 2nd screenshot that Gemini fundamentally points retired that I blindly trusted the recommendation without validating the math.

Honestly… fair.

The quality is that this correction only costs maine clone money and a Madden franchise. The Jeep illustration almost costs maine existent money, and the SEO illustration could person costs implementation momentum and trust.

The worth of expertise was ne'er memorizing answers. It was knowing erstwhile thing feels off, asking amended questions, and recognizing erstwhile the reply smells for illustration bullshit.

AI didn’t region that skill. If anything, it made it much valuable.

AI isn’t replacing experts. It’s replacing group who person stopped thinking.

Leroy2
More