Optmyzr's Co-Founder Eulogized Keywords. His Own Study Says Phrase Still Wins.

Frederick Vallaeys, Google's first AdWords Evangelist and co-founder of Optmyzr, published a piece on Search Engine Journal calling the keyword system he helped build "obsolete." His argument: keywords are becoming an advertiser artifact, AI Max is the replacement, and most teams have six months to retool. The match type study he cites tells a more cautious story.

Why this eulogy lands harder than the average obsolescence post

Every PPC blog has run the "keywords are dead" piece for two years now. This one matters because Vallaeys joined Google in 2002, ran AdWords Evangelism for a decade, and his company Optmyzr exists to optimize keyword campaigns. When the person who sold the world on keyword targeting publishes the eulogy on Search Engine Journal, that reads as an insider tell, not a hot take.

The full piece argues that synthetic keywords (intent distillations Google generates from your URL, assets, and business data) are replacing manually declared ones, and that broad match plus AI Max plus campaign-level signals is now the architecture. Vallaeys lists five things to do in the next six months: separate branded and non-branded campaigns with exact lock on branded, invest in landing pages and offline conversions, treat negatives as a final control layer, test AI Max against keyword-based campaigns with proper hold-outs, and upgrade your team from keyword picker to "intent engineer."

The shift he's describing is real on the infrastructure side. Google announced in February that DSA, automatically created assets, and campaign-level broad match all auto-upgrade to AI Max in September, and after that you cannot create new DSA campaigns through Google Ads, the Editor, or the API. That decision has been made.

The friction shows up in his own footnotes

Now look at the data he cites. Vallaeys leans on the 2026 Optmyzr State of PPC study to support the obsolescence narrative. Read the actual study and the numbers argue the other direction more than the framing wants to admit.

In non-branded accounts running both broad and exact, 74.10% of accounts saw better ROAS with exact match, with a median ROAS advantage of 100.59% over broad. Even on conversion rate, where Google's automation pitch is strongest, 62.12% of accounts still performed better with exact. Phrase match, the type the SEJ piece calls out as the survivor, drives roughly 40% of conversions with a 15.7% conversion rate. And on branded, exact match returns 6.61x ROAS at $0.90 CPC.

Those are not the numbers of a campaign type that is six months from irrelevance.

The thing Vallaeys is right about is the direction of travel. Exact match non-branded spend really did drop from 37.1% in 2022 to 27.6% in 2026, and broad really did become the single largest bucket at 38.8%. Spend is migrating. But the conversion economics inside the spend that hasn't migrated still look healthy, and pretending otherwise is the part where the editorial voice gets ahead of the data.

What the independent AI Max tests actually return

The other piece missing from the obsolescence pitch is what advertisers are getting back from AI Max in 2026. Google's official line is roughly a 7% lift in conversions or conversion value at similar CPA when the full feature suite (search term matching, text customization, and final URL expansion) is enabled. That is the marketing number from a vendor with auto-upgrade pressure on its calendar.

Independent results are messier. Altitude Marketing's 30-day test reported a small CTR uptick that didn't break out as material, and the team admitted they spent more time correcting AI-generated copy errors than manual creation would have taken. Industry coverage of early holdout results suggests roughly 84% of advertisers see neutral or negative outcomes in initial AI Max experiments. We covered a similar pattern in AI Max Holdouts Kept CPCs Flat, where DSA-heavy accounts that delayed AI Max kept CPCs stable while early adopters saw 16% CPA increases.

The honest read of AI Max in May 2026 is that it works for some account structures, fails for others, and the determining variable is signal quality more than match-type choice. That falls well short of the obsolescence claim. AI Max is a tool that beats keywords under narrow conditions, not the keyword killer the framing wants.

The half of Vallaeys's advice I'd run with this quarter

Roughly half of his five-point list holds up if you read it through the data. Half of it is six months early.

The defensive moves are correct and should ship now. Separate branded from non-branded today if you haven't, with exact match lock on branded terms. This protects the 6.61x ROAS bucket from broad match cannibalizing your own brand search, which is the single highest-margin spend in most accounts. Build out negative keyword lists more aggressively, because once DSA auto-upgrades to AI Max in September, negatives become essentially the only steering wheel left for the campaigns that get force-migrated. Audit landing pages and offline conversion feeds, because those are the signals AI Max actually reads, and weak signals are why most AI Max tests come back neutral.

The offensive moves are where I'd hedge. Vallaeys wants you to start running AI Max against your keyword campaigns with holdouts. That part is fine and, honestly, you should be doing this already. Where I would push back is the "intent engineer" reskill timeline. If 84% of advertisers are seeing neutral or worse results today, the right move is probably not to retrain your team off keywords inside six months. It is to keep the keyword craft alive on the campaigns where it is still profitable (branded exact, high-intent phrase) and add AI Max as an experimental layer rather than a wholesale replacement.

This isn't a hill to die on. The infrastructure is migrating toward synthetic keywords regardless of what any individual advertiser thinks. But the speed of the migration is being oversold by people who built the system being replaced and have books, software, and conference keynotes that depend on the next chapter selling well. The data Vallaeys himself cites is the strongest available argument for moving slower than his post suggests.

The September date that runs independent of opinion

The hard deadline is September 2026. That's when Google starts auto-upgrading remaining DSA, ACA, and campaign-level broad match campaigns to AI Max. AI Max experiments inside Google Ads are how the platform wants you to validate the change before the auto-upgrade lands, with a holdout group on the original setup and a treatment group flipped to AI Max so you can read the delta in your own account rather than someone else's blog post.

If you run DSA right now, the migration is happening to you regardless of what you think of Vallaeys's framing. The actual question is whether you let Google's defaults inherit your campaigns or whether you walk in with documented baselines from your last 30 days of DSA performance so you can detect regression after the upgrade.

I'd put a calendar event for the second week of August on that. Pull DSA performance benchmarks, set up a geo-holdout for AI Max, document negative keyword lists separately from the campaigns they live in, and tag offline conversion feeds for source quality. Then let September happen with your tripwires set instead of your fingers crossed.

Vallaeys is right that this transition is underway. He is probably wrong that you need to be six months ahead of it. The man who built the keyword system might be calling it obsolete a little early because that, more than the data, is what the next chapter of his career depends on.

Notice Me Senpai Editorial