The Answer Layer Is Inventing Robot Prices — and Signing Real Dealers' Names to Them

In July we wrote about why cobot prices are hidden and who fills the vacuum. The villains in that piece were content farms: sites with no methodology and no sources that publish confident dollar figures for machines they have never quoted, and which own the search results for "X robot price."

That problem has not gone away. But something has changed on top of it, and it is worse in one specific way: the fabricated number now arrives attributed to a real dealer, in the summary that sits above the search results, before a reader ever clicks anything.

We are not theorizing. Over a single day of ordinary price research — August 2, 2026 — we hit four instances, and every one is checkable in a browser in under a minute.

Four cases from one day

The third case deserves a second look, because of the sentence attached to it: *"This price appears to be consistent across multiple sources."* The number was wrong by 16%, and it came wrapped in a claim about corroboration.

This is not new to us, only newly frequent. Our own working notes record a robot vacuum quoted at $42,000 by a summary when the dealer page said $18,500; a gripper quoted at $5,825 and $4,999 by aggregators when the distributor's own price sheet said $9,271.41; and a robot kit quoted at $681.30 when the manufacturer's store had moved to $783.50.

Why robot prices are unusually exposed

Three conditions have to hold for a fabricated price to survive, and this market supplies all three.

First, the ground truth is missing by design. Most robotics vendors do not publish prices. When we tried to verify a dozen brands in July, the majority were quote-only end to end. A model trained to answer a question will answer it; when no authoritative number exists, it assembles one from whatever is nearby.

Second, the pages that do carry numbers carry several. This is the part the industry does to itself. On August 2 we found one dealer listing the same KEENON T10 at $13,500 in its page body and "Starting at $18,400" in that page's own header. A day earlier we found a dealer whose temi page said $3,999 in the body, "From $4,499" in its meta description — the version Google displays — and $3,495 on a second page of the same site. A system reading these pages is not choosing between right and wrong. It is choosing between three numbers the seller published simultaneously.

Third, placeholders are indistinguishable from prices. Also on August 2: one storefront listed four completely different KEENON machines — a 40 kg tray robot, a 20 kg locker robot, a floor scrubber and a 100 kg-plus industrial courier — at exactly $20,580.00 each. Another listed the German Bionic Exia at $0.01. A third listed two Universal Robots teach pendants at $0.00. None of those are prices. All of them are structured, machine-readable, and sitting in the field a scraper reads as "price."

The part that is genuinely new

Content farms were always easy to dismiss once you saw the domain. What is different now is attribution laundering: the number is delivered under the name of a real, reputable dealer that a buyer may already be talking to. There is no sketchy domain to notice. The reader's plausibility check — *is this a source I recognize?* — returns yes, and the check that would have caught it — *does that page actually say this?* — is the one nobody performs.

And this is not a robotics-specific failure. The Tow Center for Digital Journalism tested eight generative search tools in March 2025 and found they answered more than 60% of queries incorrectly; Perplexity was the best at 37% wrong, Grok-3 the worst at 94%. More than half of the citations from Gemini and Grok 3 pointed at fabricated or broken URLs. The researchers' judgment on tone is the one that matters here: the tools "presented inaccurate answers with alarming confidence, rarely using qualifying phrases," and premium tiers were *more* likely to give a definitive wrong answer rather than decline. Our four cases are that finding, landing on purchase orders.

Our view: a price is not a fact to be retrieved, it is a claim that exists only in a specific place at a specific time under specific terms. Answer engines currently treat prices as facts, and they inherit an ecosystem — hidden list prices, multi-price pages, $0.01 placeholders — that guarantees the treatment fails. Until the answer layer either quotes a price with its on-page location and date or declines to state one, the correct assumption for a buyer is that any dollar figure not read off the seller's live page is unverified. The vendors could end most of this by publishing a price. The dealers could end much of the rest by publishing one price per machine per page.

What to actually do

1. Open the page. Not the summary, not the aggregator — the seller's live page. Every case above collapses at this step.
2. Check the currency symbol on that page. Store data often returns a bare number; we have been burned by a catalog that looked like dollars and was pounds.
3. Prefer the cart price over the body copy. Where a store has a checkout, the number in the cart is the one the seller is prepared to honor. Descriptions go stale; carts less so.
4. Check the meta description too. It is frequently a different number from the body, and it is the one displayed in search results.
5. Treat a suspiciously round or repeated figure as a placeholder — especially the same price on unrelated machines.
6. Get it in writing, dated, with the model number, the term, and what is excluded. In this market that is not caution. It is the only artifact that survives contact with a purchase order.

Sources

• Columbia Journalism Review / Tow Center — "We Compared Eight AI Search Engines. They're All Bad at Citing News." — Klaudia Jaźwińska and Aisvarya Chandrasekar, March 6, 2025
• Nieman Journalism Lab — summary of the Tow Center findings
• Helping Hand Robotics — published robot pricing (read in full, August 2, 2026)
• Knoxlabs — KEENON collection
• RobotLAB — Keenon T10 and hospitality robots catalog
• Hilti USA — Construction exoskeletons
• MTN Shop — German Bionic Exia
• GrabaRobot, Why Cobot Prices Are Hidden — and Who Fills the Vacuum, July 26, 2026
• GrabaRobot, temi Robot Price Guide 2026, August 1, 2026