Fact-Check: What Seven Google Ads YouTube Lessons Get Right, Get Wrong, and Leave Out

I watched seven YouTube videos about Google Ads — one in English, six in Portuguese — and checked every claim against Google's own documentation, 2026 market benchmarks, and the internal documents Google itself was forced to hand over in the US Department of Justice antitrust case.

The result isn't what you expect from a piece about "gurus." Plenty of it is correct, more than I expected. There's also an error of two orders of magnitude presented as a case study, and there's advice that, if you follow it, has a real chance of costing you your account.

One data point worth putting on the table first: five of the seven videos are sales funnels for a course, a mentorship, or a tool. That doesn't invalidate anything on its own — people who sell courses can also operate accounts. But it explains the distribution of what I found remarkably well. The verifiable, boring parts are correct. The spectacular parts are precisely the ones that don't survive a check.

What holds up

I'll start with what's right, because that's the part nobody writes.

The auction structure described in the English video is correct. The document Google presented to the UK's CMA, entered as evidence in the DOJ case, describes Ad Rank as an LTV score built on bid, pCTR (predicted click-through), pCQ (predicted creative quality) and pLQ (predicted landing page experience), plus thresholds, reserve prices and the rGSP auction. The four stages — retrieval, elimination, ranking, auction — check out. And the claim that a negative score removes the ad before the auction also holds: there is an embedded cost to showing a bad ad, and it can flip the sign of the calculation.

Relevance between search term, ad, and landing page. It shows up in three of the seven videos and it's the market's genuine consensus. Still true, and still the first thing to fix in almost every account I open.

Feed-only Performance Max exists and works. You create the asset group pointing at the feed URLs and leave the creative fields blank. Google complains, throws warnings, and the group saves with "Limited" status — which is exactly the outcome you wanted. Serious agencies use this.

The no-impressions checklist. One of the videos lists eight causes: scheduling dates, inactive account, bid too low, insufficient budget, no search volume, match type too restrictive, overlapping targeting, and a geographic radius that's too small. These are verifiable checks inside your own account, with no promised outcome attached. It's the most honest piece of the whole set, and the only one I'd recommend to someone starting out.

Where the numbers don't add up

The CTR-by-position chart is organic sold as paid

One of the videos shows the classic distribution: 40% of clicks on position 1, 20% on 2, 10% on 3, 7.2% on 4. The numbers are right. The context is wrong. Those percentages come from organic search studies — position 1 around 39.8%, position 2 at 18.7%, position 3 at 10.2%.

For ads, the order of magnitude is different. The paid top slot sits near 2.1% CTR, and the first organic result gets on average nineteen times more clicks than the first ad. The video reads an organic chart and concludes that "position 1 takes 40% of the money" in paid media. It doesn't. And the entire strategy built on top of that reading — including the most dangerous one of all, which I'll get to — comes from a mislabeled axis.

Quality Score is not Ad Rank

Another video treats the 1-10 Quality Score as if it were the quality component of the auction. It isn't. The number you see in that column is diagnostic: it tells you where the problem is, it doesn't enter the calculation. Ad Rank has six inputs and is recalculated in real time, in every auction, with contextual signals that never appear in that column.

The practical consequence of the confusion is people chasing 10/10 as if it were a campaign goal. Quality Score is a thermometer, not a KPI.

The 48 customers who don't exist

The Display Network video presents a remarketing case: 97 visitors, 48 sales, R$29,000 in revenue. That's practically a 50% conversion rate.

The Display conversion benchmark sits near 0.57%. Retargeting CTR on Google runs between 0.7% and 1.2% — CTR, not conversion. The case presented is two orders of magnitude above what the network delivers.

This isn't a "every account is different" situation. It's the difference between 0.5% and 50%. A number like that only appears through a tiny sample cherry-picked after the fact, contaminated attribution, or invention. None of those three hypotheses justifies you building an operation on top of it.

"Redirecting is a crime under each country's laws"

False. Automatic redirection to another domain is ad policy, not criminal law. Google disapproves when the final URL or the tracking template sends the user to a domain other than the declared one, and the guidance is to remove the redirect.

The catch is that the video's practical conclusion is right: you do need your own structure, a page that is genuinely yours. It's only the justification that's invented. And that matters, because someone who learns a rule through the wrong justification can't recognize the neighboring cases — like destination mismatch, which is the same family of disapproval and has nothing to do with legislation.

Automated CPA on a brand-new account

One of the videos sells Target CPA from day one, "with contingency." The operational threshold is roughly 30 conversions in 30 days for Smart Bidding to beat manual bidding. Below that, the model doesn't have enough signal to model anything. For tROAS, the practical floor operators cite is on the order of 50 conversions in 30 days.

New account, new product, small budget: you don't have that volume. Turning on automated CPA there isn't an aggressive strategy, it's handing the decision to a blind model.

Feed-only changed since the recording

Two corrections. Today you can no longer save an asset group with no assets at all through the interface — at least three headlines are required. And the resulting inventory is Shopping + Display, not "search and shopping" as the video claims.

It's the most forgivable error on the list: it was true when it was recorded. But it's also exactly why following a Google Ads tutorial without checking the date is an expensive way to learn.

What none of the seven mentions

Here's the part that made me write this article.

Every one of the videos, without exception, tells the same story about the auction: whoever offers the best experience wins, money doesn't buy position, relevance is king. It's a good narrative. It's also Google's official narrative, repeated without anyone going to read what came out of the antitrust case.

The internal documents revealed at trial describe mechanisms that appear in no public manual:

Project Momiji (2017): artificially inflated the second-place bid, which resulted in roughly 15% more paid by the auction winner.
Squashing: manipulated the second-place advertiser's pCTR to push the price upward, without that ad actually being more relevant.

Notice what that means. The price you pay isn't purely a function of your quality and the quality of whoever is bidding against you. In at least two documented periods, it was also a function of an adjustment made to lift revenue — inside an auction presented to the market as neutral.

This doesn't mean optimization is pointless. It means "whoever delivers the better experience pays less" is a half-truth told by a party with an interest in the omitted half. And it means none of the seven creators went to the primary source before repeating what Google says about itself — the exact same kind of uncritical reading they charge their students for doing about competitors.

The three recommendations that can cost you your account

This is the section I'd recommend reading even if you never watch any of the videos.

Duplicating the same product across five or six stores to occupy positions 1 through 6. This is the "surface area scaling" idea from the English video, and it's a direct descendant of the mislabeled CTR chart above. It's also a straight violation of Abusing the ad network: you can't run ads for the same site from two accounts, and duplicating the site across distinct domains to get around the rule is explicitly named as circumvention. Consequence: disapproval, suspension, ban.

"I have several accounts, all of them blocked, I need to create more." This shows up casually in two videos, said almost as a badge of experience. It's Circumventing Systems — creating a new account to get back in after a suspension is the literal definition of the policy. Consequence: another suspension, and risk of a permanent ban on the associated identity documents and payment methods.

One honest caveat: having multiple accounts for legitimate operational reasons — agencies, separate teams, different regions, client structures — is not a violation. The problem was never the number of accounts. It's creating an account to get around a suspension, or to occupy more ad space than you're entitled to.

A presell page with the affiliate link straight on the button. It reduces destination-mismatch risk, that much is true. But it doesn't cover low-value content or offer misrepresentation — the two disapprovals that actually take affiliate pages down. Calling that "being safe" is selling half a solution as armor.

None of the seven videos flags any of these risks. Not even in a footnote.

What's left once you strip the noise

Not much is left, and what's left is good. This is the usable consensus across all seven videos, and it's what I'd sign my name to:

Consistency between search term, creative, and landing page. It always was this, and it still is.
Quality Score as a diagnostic, not a goal. Don't chase 10/10.
Bids and budgets proportional to the keyword's real CPC. Opening Keyword Planner before raising a bid solves half of all "I'm not getting impressions" cases.
Continuous negative keywords built from the search terms report. It's manual, weekly, glamour-free work — and it's where the money is.
Don't pause a campaign before you have enough data. Adjust bids on what's spending the most instead of pausing everything and starting over, because starting over costs learning.
Over-segmentation chokes delivery. Age plus gender plus income plus a three-mile radius is a funnel so narrow no impression gets through.

Six lines. That's honestly what seven videos, added together, deliver in usable material — and all six are free, verifiable, and boring. You can't sell a course with that.

The detail almost nobody checks: benchmarks have context

One point worth calling out, because it's where I see competent people go wrong most often.

Nearly all of this material is e-commerce and affiliate: short purchase cycle, conversion happening on the page itself, impulse products. If your operation is something else — local lead generation, a long-cycle service, B2B, rentals, healthcare, education — three things don't transfer:

Conversion benchmarks. Local lead generation converts far above e-commerce. Using these videos' numbers as your reference makes you think you're doing badly when you're doing well, or the reverse.
"Duplicate to occupy slots." Inapplicable with a single brand, and risky in any configuration.
Cold Display as an entry channel. The low CPC doesn't compensate for the low intent outside an impulse-product context. In services, you're buying click volume that never becomes a conversation.

What does transfer well: the no-impressions diagnostic checklist and the routine of reading search terms and adding negatives. Note that those are precisely the two most operational and least sellable parts of the whole set.

How to watch the next free lesson

The lesson isn't "don't watch." It's that free material has a predictable incentive structure, and knowing that structure lets you take the good parts without swallowing the rest.

The rule I use: the most spectacular claim in the video is the one carrying the sale. It isn't a coincidence that the two-orders-of-magnitude error sits exactly in the results case study and not in the diagnostic checklist. The checklist sells nothing. The case study sells everything. Which is why the checklist got checked and the case study didn't.

Three questions come out of that, thirty seconds each:

What's the primary source for this number? If the chart has no label saying whether it's organic or paid, it doesn't count as an argument.

Is this tactic allowed by the platform? Not by law, not by common sense: by the written policy. Searching the English name of the practice — double serving, circumventing systems, bridge page — settles it on the first page of results.

When was this recorded? Feed-only changed in months. Smart Bidding changed in August. A paid media tutorial has a shorter shelf life than YouTube's willingness to recommend it.

Three questions. They don't make you cynical — they make you capable of using free lessons without outsourcing your judgment. Which, in the end, is the only thing separating people who learn from internet content from people who repeat internet content.

If you want someone to go through your account with that same level of checking — what's right, what's wrong, and what could take you down — that's exactly the work I do. Get in touch.

References

— views— likes