Deeper Diligence

Ad Fraud, Bots & AI Agents: How to Stop Paying for Fake Traffic (w/ Andrew Barrow, RevenueArc)

On Deeper Diligence, host Bogdan Patynski and co-host Jeromy Sonne (CEO of Daypart) talk with Andrew Barrow, former holdco business development leader and founder of RevenueArc, about the under-discussed reality of ad fraud, bots, verification, and why AI is making the problem worse. Barrow explains how bots now mimic real users by interacting with ads via virtual machines, argues against the idea of advertising to AI agents because major LLMs are instructed to ignore ads, and recommends focusing on GEO/AEO and being a trusted reference source instead. He describes RevenueArc’s approach to avoiding fraud and brand-safety risks by skipping the open web, buying only PMPs, manually vetting sites and apps, and using targeted propensity modeling (Arc Foresight). He advises practitioners to turn off open exchanges, extended networks, and Google PMax, expect lower traffic but better quality, and test tools by watching for reduced bounce rates and improved performance.

00:00 Ad Spend Gone Wrong

00:20 Meet the Hosts

00:40 Guest Intro RevenueArc

01:45 Fraud in Advertising

02:06 How Bots Hide

03:22 Agents Versus Bots

04:11 Why LLMs Ignore Ads

06:55 Avoiding Open Web

08:13 Should Agents See Ads

12:18 Bot Fraud Origins

13:50 RevenueArc Playbook

15:32 Arc Foresight Targeting

16:42 Quick Fixes Today

18:04 Testing Future Solutions

19:41 Contact and Wrap Up

This podcast is powered by Pinecast.

Host Bogdan Patynski and co-host Digital_Chadvertising interview Heather Physioc, Chief Discoverability Officer at VML, about credibility, verification, and brand discoverability across search and AI-driven platforms.

Heather talks about credibility which is cumulative and outlines three early steps: verify identity via real human signals and authoritative connections, publish high-quality original work with a clear point of view, and earn third-party validation.

They discuss fake audiences, temporary verification badges, black-hat tactics, and how scaling AI-generated content can backfire by looking spammy and damaging trust.

The conversation covers repetition vs corroboration, concerns about Google AI Overviews accuracy, emerging liability (including a Munich court ruling holding Google responsible for false AI overview claims), declining organic traffic, AI shaping brand narratives using off-site sources, the rise of junk sites, and a shift from open-web search to “cozier communities” like Reddit and Substack. Heather shares that LinkedIn is the easiest way to reach her.

00:00 Welcome And Introductions

00:56 March Chadness Explained

01:40 Building Credibility Fast

04:29 Fake Followers And Search Spam

05:21 Meaningful Verification Signals

07:27 Black Hat SEO And AI Content

10:22 Spotting Guru Credibility

14:03 AI Overviews And Misinformation

15:19 Who Is Liable For AI

18:38 Brand Search In The AI Era

20:54 Junk Sites And Vibe Coding

22:00 Incentives Investors And Ads

24:27 Cozy Communities Replace Search

27:11 Where To Find Heather

This podcast is powered by Pinecast.

In a Deeper Diligence guest episode, privacy researcher and former FTC researcher Shoshana Wodinsky explains what advertisers actually know about consumers—and how much of the data economy is built on inference rather than fact.

Wodinsky breaks down how browsing behavior, purchases, device IDs, location signals, Wi-Fi logins, and other data points become audience profiles, including how a segment supposedly identifying 80,000 pregnant women ultimately traced back to devices that had simply passed near a Babies R Us.

She also discusses behavioral targeting in sports betting, why AI may scale questionable advertising inferences rather than solve them, and a case where device and location data were used to identify and track a Catholic priest. The conversation examines the difference between knowing and inferring, why supposedly anonymous data can still identify people, what consumers can realistically do to protect themselves, and why meaningful consent remains difficult in modern advertising.

Subscribe to The Diligence Report:

https://newsletter.daypart.ai/

Follow Shoshana Wodinsky:

LinkedIn: https://www.linkedin.com/in/swodinsky

X / Twitter: https://x.com/swodinsky

00:00 Welcome and Guest Intro

01:19 What Advertisers Actually Know About You

02:24 The “80,000 Pregnant Women” Audience

03:34 How Advertisers Infer Your Location

04:55 Gambling and Predatory Targeting

05:51 How Sports Betting Sites Nudge Behavior

07:05 Does AI Make Ad Targeting Better?

07:27 “Whoever Has the Most Data Wins”

08:35 When Advertising Inferences Get It Wrong

09:26 Why Targeted Ads Can Still Miss

10:12 Why Big Platforms Are Hard to Investigate

11:30 How Ad Tech Was Used to Identify a Priest

12:20 How “Anonymous” Data Can Reveal Your Identity

13:24 The Real-World Consequences of Location Data

14:16 Can You Actually Protect Your Data?

15:19 What Privacy Policies Reveal

16:22 Why Consumers Can’t Solve Privacy Alone

17:24 Can Targeted Advertising Exist Without Surveillance?

17:43 Consent and “No Means No”

18:26 Closing and Where to Find Shoshana

This podcast is powered by Pinecast.

Links:

newsletter.daypart.ai

robertfreundlaw.com

Robert Freund, an advertising and e-commerce lawyer, explains that most AI advertising risk comes from applying longstanding FTC Act Section 5 truth-in-advertising principles and endorsement rules rather than new federal AI-specific laws.

He highlights emerging state disclosure requirements for AI-created “synthetic performers,” including New York’s law (with AG-only enforcement and stated penalties of $1,000 for a first violation and $5,000 for subsequent violations), Hawaii’s similar law, and a pending California version.

He warns that disclosures don’t cure underlying deception, such as AI-generated before/after images or AI avatars giving “experience-based” testimonials, and notes deepfakes can trigger publicity rights and copyright issues. Responsibility generally remains with the advertiser and other parties who materially contribute, even when platforms generate creatives.

Recommended practices include identifying applicable risks, strengthening e-commerce terms of sale (including arbitration/class action waiver), negotiating indemnities/limits with vendors, and auditing content before scaling.

00:00 Welcome and Guest Intro

00:10 Rob’s Legal Background

01:19 AI Ads and Existing FTC Rules

04:20 State AI Disclosure Laws

08:27 Will Enforcement Expand

10:51 Who Gets Targeted

14:24 Creative Liability Checklist

18:36 Platform Tools and Shared Liability

21:33 Penalties and Damage Ranges

25:35 Compliance Best Practices

27:35 Where to Follow Rob

This podcast is powered by Pinecast.

Bogdan Patynski and Digital Chadvertising hosts Deeper Diligence with Steven Liss, co-founder and CEO of OpenAds.ai, discussing “ads for agents,” where sponsored information is served to AI crawlers (e.g., ChatGPT, Claude, Gemini) and can surface in user answers.

Steven explains the market pull from brands wanting visibility in AI results, publishers’ monetization pressure as referral traffic drops, and why ads-to-agents could be an alternative to pay-per-scrape models.

We cover Perplexity’s stance on blocking such ads, likely responses from major labs, and publisher counter-leverage. Steven also shares experiments using CDN detection to serve different agent-specific content, emphasizing that success depends on retrieval/SEO and that performance measurement can be done via referral codes and cost-per-action pricing. We also discuss disclosure, standards, gray-hat manipulation risks, and why legacy DSPs struggle with purely contextual, non-identity targeting.

00:00 Show Premise

00:27 Meet Steven Liss

01:11 Ads For Agents

01:58 Why Now

03:20 Platform Pushback

04:29 Who Pays

05:15 Blocking And Evasion

06:29 Trust And Relevance

08:10 Experiment Setup

09:58 Prompt Sensitivity

11:53 Agent Autonomy

12:51 Measuring With Codes

13:34 B2B Opportunity

14:30 Programmatic Future

15:28 DSP Challenges

17:42 Pricing And CPA

18:36 Power Dynamics

22:46 Standards And Disclosure

23:48 Gray Hat Risks

25:37 DSPs And Budgets

28:46 Fraud Watchouts

30:15 Agent Internet Future

31:05 Where To Follow

This podcast is powered by Pinecast.

Bogdan and co-host Digital Chatvertising interview Dr. Augustine Fou of Fou Analytics, a longtime digital marketer and ad-fraud researcher (ex‑McKinsey, American Express, Interpublic, Omnicom; PhD MIT), on why pixel attribution dominates because it’s easy and bundled with platforms, yet is oversimplistic and easily gamed. Fou explains better validation via holdout/“turn‑off” experiments (citing eBay’s paid-search test) to measure incremental impact using traffic and sales velocity, and recommends repeating on/off cycles for causal evidence. The discussion covers fraud seasonality spikes around budget deadlines (Q4, month/quarter ends) and in movie and political spending, how programmatic exchanges enabled fraud via fake sites and low-value apps, and why AI mainly lowers barriers rather than creating a new fraud wave. He details attribution gaming (e.g., footfall credit via mass device exposure) and advises reconciling platform-attributed conversions against real sales. Fou argues for evidence-based verification and transparent supporting data, contrasting legacy verification approaches, and offers common-sense ways to spot MFA sites.

00:00 Welcome and Introductions

01:27 Why Pixel Attribution Wins

04:25 Beyond Pixels Holdout Tests

07:01 Budget Pressure and Incrementality

08:57 Fraud Seasonality in Q4

11:51 Who Are the Bad Guys

16:24 Old Fraud Still Thrives

18:05 AI and Modern Fraud Tricks

20:01 How Attribution Gets Gamed

25:17 DIY Truth Checks for Nontechnical Teams

28:58 Turn Off Experiment Playbook

30:19 Verify Traffic and Vendor Claims

37:04 Spotting MFA Sites Fast

40:04 Wrap Up and Where to Reach Dr Fou

This podcast is powered by Pinecast.

In this episode of Deeper Diligence, Digital_Chadvertising and guest Eric Tilbury (VP of Programmatic & Solutions Engineering at Inuvo) break down how ad attribution works and why it often fails. They explain user-ID and pixel-based attribution, including view-through vs click-through conversions, and discuss alternative approaches like media mix modeling and incrementality testing (e.g., geo lift). The conversation covers how privacy changes (like Safari blocking third-party trackers) and cross-device behavior make deterministic attribution fragile, while black-box platforms and incentives to hit low CPA/ROAS can distort results. They discuss common gaming and fraud tactics such as ID bridging, affiliate and click manipulation, and bottom-funnel retargeting that captures outsized credit, leading to budget misallocation. Practical advice includes defining measurement strategy first, testing vendors, building in-house measurement protocols, and using transparency and independent verification to evaluate ad tech.

00:00 Welcome and Guest Intro

01:51 What Attribution Means

03:27 View Through vs Click Through

04:16 Pixels and Tracking Basics

05:43 Other Attribution Methods

07:50 When Attribution Breaks

08:59 Incrementality and MMM

11:50 Why Deterministic Misleads

14:39 Easy Button Incentives

18:42 Privacy and ID Limits

22:17 ID Bridging and Fraud

28:08 Retargeting Reality Check

31:42 CFO and Investor Playbook

33:36 Transparency and Vetting

35:26 Future of Measurement

37:03 Wrap Up and Where to Find Eric

This podcast is powered by Pinecast.

Jeromy, Bogdan and Lee discuss Google integrating AI more deeply into its ad campaign workflow while changing terms so advertisers are liable for AI-generated mistakes, arguing Google gains control without responsibility and predicting potential lawsuits and backlash, especially from professional advertisers. They explore why Google may be pushing this, including internal incentives to grow YouTube advertiser counts, and argue better measurement tools for incremental lift would drive adoption more sustainably. The conversation shifts to an Australian dock workers’ union demanding a 28-hour workweek as automation and AI increase productivity, questioning whether claimed gains are real and where they actually occur. They then cover a Texas Tribune investigation into Texas data centers using permitting tactics for turbines and diesel generators, noting emissions tied to gas plants serving data centers and debating hype versus reality of build-outs. Finally, they discuss Apple reportedly considering acquiring Prism ML to run a compressed 27B-parameter model locally on iPhones, emphasizing privacy and potential market shifts if Apple executes.

00:00 Welcome and Introductions

01:08 Google Ads AI Liability

06:27 Why Google Pushes YouTube

09:20 Measuring CTV Lift Properly

12:09 Australia 28 Hour Workweek

13:14 AI Productivity Reality Check

21:28 Texas Data Centers Emissions

27:59 North Korea Linux Tangent

30:42 Apple Local AI Comeback

37:00 Wrap Up and Subscribe

This podcast is powered by Pinecast.

Ben Edelman on Phia’s Alleged Forced Clicks: How Shopping Plugins Commit Affiliate Fraud

In a Deeper Diligence guest episode, ad-fraud researcher Ben Edelman discusses his investigation into Phia and the Bloomberg coverage alleging affiliate fraud through “forced clicks” and standdown-rule violations. Edelman explains the three-step affiliate marketing bargain (show link, user clicks, user buys) and how forced clicks skip the user click by using Phia’s iOS plugin to open an invisible tab that loads an affiliate link and closes it, positioning Phia as “last click” for commission—often costing merchants, and sometimes other affiliates like review publishers. He also describes standdown rules requiring shopping plugins to stay out of the way when another affiliate referred the user, and says Phia tracked competitor affiliate links yet did not stand down. Edelman questions Phia’s claim this behavior was a bug, cites past cases including eBay prosecutions and Honey-related evidence, and argues enforcement can deter fraud.

00:00 Welcome and Guest Intro

01:14 Ben Edelman Background

01:40 Why Phia Drew Attention

03:18 Cookie Stuffing vs Forced Clicks

04:14 Affiliate Marketing Basics

07:21 Networks and Incentives

09:40 Phia Forced Clicks Explained

12:34 Stand Down Rules Violations

14:31 Real World Harm Examples

17:09 Bug Claim and Intent

21:27 Past Major Fraud Cases

23:27 Honey Investigation Lessons

24:55 Stopping Fraud Systemically

27:02 Future Misconduct and Wrap Up

28:11 Closing Thanks and Contact

This podcast is powered by Pinecast.