Day 86: Is Nextdoor Laboring? Or Are We Back to Business as Usual?

Labor Day is over.

It's Day 86 of my continuing Nextdoor case study, and after wondering yesterday whether Nextdoor was taking an extended holiday weekend, I checked again today.

I haven't received any communication regarding the two detailed Insights studies I requested.

And NXDR?

As of this writing, it's hovering around the low $2.20s. Market sources were showing roughly $2.24 late this afternoon after Friday's $2.24 close, so the exact number is moving as I write this.

Hey, at least we're talking pennies now.

Is Nextdoor laboring?

I'm still trying to figure that out.

But something Nirav Tolia posted on X on September 3 caught my attention.

“A 2025 Study Found...”

In the post, Nirav wrote:

“A 2025 study found AI-generated fake reviews are now indistinguishable from real ones...”

And immediately I thought:

Here we go again.

A study.

Which study?

Who wrote it?

Where was it published?

What was the methodology?

What was the sample size?

What exactly did “indistinguishable” mean?

This probably sounds familiar to anyone who has followed my 86-day adventure in Nextdoor research transparency.

Here's the interesting part: I looked, and there is identifiable research behind Nirav's statement.

In a Fortune piece published the same day, Nirav identified it more specifically as a 2025 study from Nottingham University Business School.

Peer-reviewed 2025 studies examine AI-generated fake reviews, including a Journal of Retailing and Consumer Services study analyzing 714,016 reviews and finding meaningful linguistic differences between AI-generated fake reviews, human-generated fake reviews, and authentic reviews.

So my criticism isn't:

“The study doesn't exist.”

My question is:

Why not cite it in the original post?

If you're going to use research to establish credibility, give readers enough information to evaluate the research themselves.

Study name. Authors. Publication. Link.

Four things.

Done.

Especially when you're the CEO of a company currently promoting the idea that provenance matters.

That's almost too perfect.

Fake Reviews Didn't Arrive With ChatGPT

Another part of this discussion bothers me.

AI didn't invent fake reviews.

AI made them faster, cheaper, and easier to produce at scale.

Humans have been manipulating reputations for a very, very long time.

In fact, the history is pretty entertaining.

1800s: Walt Whitman anonymously published glowing reviews of his own Leaves of Grass. That's essentially the 19th-century version of creating a burner account and giving yourself five stars.

Early 1900s: Newspapers were already dealing with fabricated information, exaggerated claims, and “fakers.” By 1913, the New York World had established a Bureau of Accuracy and Fair Play partly to address complaints and “stamp out fakes and fakers.”

Pre-internet advertising: Businesses used testimonials and endorsements as marketing tools, creating the same fundamental problem we face today: Is this person recommending the product because they genuinely love it—or because somebody benefits from the endorsement?

Early Internet: Message boards, review sites, and eventually e-commerce let businesses and individuals create accounts and manufacture praise—or attack competitors.

Ironically, Nirav himself co-founded Epinions in 1999, one of the early user-generated review platforms. He now acknowledges that the review economy eventually became vulnerable to manipulation.

2010s: Fake-review businesses became an industry. Businesses could purchase positive reviews, competitors could be attacked with negative ones, and reviewers could operate multiple identities.

2020s: Review farms became increasingly sophisticated. Investigators found coordinated reviewers praising unrelated businesses across countries, with identical reviews sometimes appearing under different names.

Today: Generative AI has dramatically changed the economics. Instead of paying humans to crank out hundreds of reviews manually, someone can rapidly generate enormous quantities of convincing text. Research now shows LLMs can produce deceptive reviews with human-level capabilities, even though researchers can still statistically identify some linguistic differences.

Same scam.

Much better machinery.

Which Brings Me Back to Nextdoor

Nextdoor's current pitch is that verified identity and neighborhood accountability can make recommendations more trustworthy.

That's an interesting strategy.

But verification alone doesn't make an opinion truthful.

A real person can exaggerate.

A verified person can have a grudge.

A real neighbor can recommend their friend's company.

A legitimate customer can receive an incentive.

Two verified neighbors can have completely different experiences with the same business.

And, as I've previously documented, I've had my own questions about how robust Nextdoor's verification system actually is.

So “verified” is a trust signal.

It isn't a magical truth serum.

Provenance Matters? I agree.

That's actually what makes Nirav's post so interesting.

His broader argument is essentially:

Where information comes from matters.

I agree.

Completely.

Which is why when the CEO says:

“A 2025 study found...”

I want the provenance.

Name the study.

Name the researchers.

Link to it.

Let readers inspect it.

And that's precisely what I've been asking Nextdoor to provide regarding its own research.

For 86 days.

Nextdoor published Insights findings.

Nextdoor publicly told readers they could request detailed data.

I requested it.

I've emailed.

I've followed up.

I've expanded the distribution.

I sent another request Saturday.

And I'm still waiting for the two detailed studies.

Day 86

Yesterday was Labor Day.

Today I'm asking:

Is Nextdoor laboring?

Because I'm still doing the work.

Waiting for somebody at Nextdoor to answer the questions.

Nirav's post argues that provenance matters.

On that, we agree.

So here's my Day 86 suggestion:

Practice it.

When you cite research, show us the research.

When Nextdoor publishes research, show us the methodology.

When your own blog tells readers they can request detailed data, provide the detailed data—or explain why you won't.

Because whether we're talking about a restaurant review, an AI-generated recommendation, or a corporate Insights report, the principle is the same:

Don't just tell me to trust it. Give me enough information to decide whether to trust it.

Follow the continuing Nextdoor case study at NielFlamm.com/blog.