Sulat.com / Writing on AISunday, November 22, 2026
Big Tech/22 November 2026

Andrew Shotland says a Google AI Overview gave him a Vanguard number that led to a scam call

Search expert Andrew Shotland says a number from a Google AI Overview put him through to scammers posing as Vanguard. His column doesn't show the number, and it isn't the first report of its kind.

Glossy 3D Google G logo beside a green phone handset, a warning speech bubble and a magnifying glass

At a glance

  • Andrew Shotland, CEO of Local SEO Guide, wrote in Search Engine Land on 20 November that a number from a Google AI Overview connected him to scammers while he was trying to reach Vanguard.
  • He says the caller claimed his Social Security number was exposed, and that Vanguard and Bank of America both had no record of the call the next morning.
  • His column doesn't include the number, a screenshot of the Overview or the words he searched, and it carries no comment from Google or Vanguard.
  • Reporters have documented fake support numbers in Google AI Overviews since August 2025, and Google has said it takes action on specific examples.
  • The check is the same either way: get a support number from your card, a statement or the official app, not from an AI answer.

Andrew Shotland runs a local search consultancy, and he says he still called a scammer. In a Search Engine Land column published on 20 November 2026, the CEO of Local SEO Guide writes that he searched Google for Vanguard’s customer support number, called the number an AI Overview showed him, and gave information to a caller who, by his account, wasn’t Vanguard.

Glenn Gabe, an SEO and AI search consultant, shared the column on X on 21 November and told his followers not to trust what they see in an AI Overview or any AI response. Everything below about the call comes from Shotland’s own account, and none of it has been independently checked.

Glenn Gabe’s post of 21 November 2026, quoting Shotland’s column.

What Shotland says happened

Shotland writes that he searched for Vanguard’s customer support number the day before the column ran. Google’s AI Overview gave him a number, and he called it “without thinking twice.”

The man who answered “seemed legit,” he writes. He asked for verification information, paused, and put Shotland on hold. When he came back, he said Shotland’s Social Security number and other personal information had been compromised and were “all over the dark web.” He then asked whether Shotland had traded crypto and whether he’d committed serious crimes, and followed each answer by asking if he was certain.

Asked where else he banked, Shotland said Bank of America. The caller put him on hold to speak with the bank, came back, and said a call from BofA Fraud Detection would follow. Seconds later, “Bank of America” showed on Shotland’s caller ID. “I completely fell for it,” he writes. “I gave these guys way too much information.”

Five-step chain: searched, called, told data exposed, put on hold, no record at Vanguard or BofA next morning
Shotland’s account runs from a search to a call neither firm had on record.

The next morning Shotland called Vanguard himself and “learned they had no record of my call. Neither did BofA.” He says he froze all his accounts before anything was taken.

Don’t call customer support numbers from AI.

Andrew Shotland, Search Engine Land, 20 November 2026
Why did “Bank of America” show on the caller ID?

Search Engine Land’s summary of the column lists spoofed caller IDs among the tricks that made the call convincing. The column doesn’t say how the display was produced.

What the column shows and what it doesn’t

Shotland doesn’t print the number he called. The column has no screenshot of the AI Overview, doesn’t give the words he searched, and doesn’t say which page the Overview drew the number from. It carries no comment from Google or Vanguard, and I found no public statement from either about his account by 11pm NZDT on 22 November.

Two columns: what Shotland's piece includes versus the number, screenshot, search words and any comment it leaves out
The piece gives an account but not the number, a screenshot or a Google comment.

So nobody outside Shotland can check which number he dialled or what the Overview displayed. What the column establishes is his account of the call and his statement that Vanguard and Bank of America had no record of it. That’s why no number is named here as a scam line.

It isn’t the first report

Scam support numbers in Google’s AI answers have been reported since August 2025. Each report below is a reporter’s, victim’s or researcher’s account.

ReportedSourceWhat it said
15 August 2025The Washington PostAlex Rivlin said a Royal Caribbean number from a Google AI Overview reached scammers. A Google spokesperson said the company had “taken action” on several fake numbers.
17 August 2025Digital TrendsReddit reports, including a Southwest Airlines number that the post says came from an AI Overview and isn’t on Southwest’s website.
8 December 2025AurascapeResearch on planted support numbers. It says an Emirates query returned two numbers in a Google AI Overview that aren’t Emirates’.
15 February 2026WiredAn explainer that says not to believe an AI Overview “especially not when it comes to specific facts, figures, or phone numbers.”
Earlier reports of scam support numbers in Google AI Overviews, as the outlets described them.

US$768

What Rivlin agreed to pay for a shuttle, TechSpot reported. He cancelled his card and the charges were reversed.

2

Numbers in one Emirates answer that Aurascape says aren’t Emirates’.

15 Aug 2025

When The Washington Post reported the Royal Caribbean case.

What Google has said

In August 2025 a Google spokesperson told The Washington Post the company had “taken action” on several fake numbers and continues to work on improvements. In a statement to ZDNET, Google said:

We have strong protections and policies to prevent scams from appearing in AI Overviews or ranking highly on Search, and our systems are effective at surfacing official customer service information for the queries people search most. We’ve taken action on several of the examples shared, and continue to work on broader improvements to address rarer queries like these.

Google statement to ZDNET, August 2025

Google’s Search Help page for AI Overviews says they “can and will make mistakes” and tells people: “Always check important info in more than one place.” In February 2026 the Indian Express reported that Google said it had added safeguards to improve scam detection in AI Overviews.

How a bad number gets into an answer

Aurascape, a security firm, published research on 8 December 2025 describing what it says is a campaign to plant fake support numbers where AI systems will find them. Its researchers say attackers seeded compromised high-authority websites, including government, university and WordPress sites, and used YouTube and Yelp to host text pairing a brand name with a number in an easy-to-parse form. Answer systems then merged those sources into one confident answer.

Four-step flow: scammer plants a number, answer systems read it, merge it into one answer, caller dials it
Aurascape’s research describes how a planted number reaches a single confident answer.

Aurascape tested Perplexity and Google’s AI Overview. For an Emirates query, it says the Overview gave booking instructions with two numbers that aren’t Emirates’. Digital Trends and ZDNET also pointed to research from Odin showing that prompt injection can make Gemini include a fake support number in a summary.

Manipulating these new answer engines using techniques from 30 years ago is like shooting sitting ducks.

Mike Blumenthal, Near Media, quoted by TechSpot

The format matters too. ZDNET’s explanation is that a traditional search shows a list, most of it legitimate, while an AI answer gives one result, “increasing the odds that you’ll follow only that one.”

How to get a real number

The fix doesn’t depend on whether any particular Overview is right. Treat a number from an AI answer as unchecked, and get the real one from something you already hold.

  1. Use the number on the back of your card, or on a recent statement or bill.
  2. Open the firm’s own app, or type its web address yourself and read the Contact page.
  3. If a caller says your accounts are at risk, hang up and call back on a number from step 1 or 2.
  4. Treat a matching caller ID as no proof, since the display isn’t reliable.
Four icons: back of your card, latest statement, official app and an address you type yourself
Get the number from something you already hold, not from an AI answer.

Investors who receive a suspicious communication from a financial institution should contact the institution separately using verified public contact information to raise their concerns and report the correspondence.

Vanguard, “Spotting and stopping more sophisticated scams,” 2025

Vanguard’s US guidance also says it will never ask investors to perform a transaction to “protect” their assets, or to keep activities secret from their trusted contacts.


AI Overview
The AI-written summary Google shows above some search results. Google’s own help page says it can make mistakes.
Search poisoning
Planting content so a search or answer system surfaces it, here a fake number paired with a real brand name.
Your Money or Your Life
The label Google’s search quality guidelines use for topics that can affect a person’s finances, health or safety. Shotland’s headline plays on it.
Related: Glenn Gabe on Google’s Connected Apps in AI Mode.

Sources: Andrew Shotland, Search Engine Land; Glenn Gabe on X; The Washington Post; Digital Trends; ZDNET; TechSpot; Aurascape; Wired; Indian Express; Google Search Help; Vanguard.


Discover more from Sulat AI & Tech News

Subscribe to get the latest posts sent to your email.