Keywords are dead


Spend a few minutes on LinkedIn and you’re sure to see some frustrated PPC agency owner showing how Google’s AI is matching their [exact match] keywords to some “grossly unrelated” search query.

This comes as no surprise to those of us adopting modern search techniques, as keyword matching was updated a few years ago. Paid search practices have gone from the use of tightly controlled keywords to targeting search queries that have the same meaning as your keyword.

This has evolved largely because the way people are searching has drastically changed. We can now search using images, speech, and conversations thanks to AI.

Add this progression to the advancements that Google’s AI has made over the years as it’s gotten significantly better at understanding the actual intent behind a user’s query. It can decipher what a user is actually looking for even though what a user typed in may differ.

So marketers who are trying to use outdated methods (match-types, CPC limits, SKAGs, bid modifiers etc…) to control which search queries their ads show up for are doing themselves (and their clients) a major disservice.

Not only are they missing out on a ton of potential profitable sales volume, but they’re also missing the opportunity to train this insanely good advertising algorithm to find more pockets of demand.

Cherry picking specific keywords is boring, expensive, and fishing in a body of water that is starting to dry up.

You want Google to be able to find a bunch of new queries for you to tap into. That’s where you can find explosive growth.

At its simplest level, Google is trying to match queries to meet objectives. And it’s your job to teach Google as much as you can about your objective and how to get there.

Since keywords are no longer a targeting mechanism, you’ve got to find ways to tell Google what types of traffic and results you want from your investment.

Fail to do this and you’re going to end up with poor results... and complaining on LinkedIn.

The most important “thing” you can tell Google is of course how you make money by using your conversion pixel — which fires when a user completes a given event along with an associated value of that unique event (usually purchases but sometimes a form-submit, page view, etc..).

This conversion objective and the value that comes with it is what powers value-based bidding (tROAS) which is an easy example of how your inputs result in outputs from Google.

However conversion tracking is table stakes at this point…

The real way to train the algorithm to get you more of what you want is through signals & theming.

This is evidenced by the latest rollout of Search Themes into Performance Max (which will be the future of Paid Search whether you like it or not), which Google describes as follows:

Let’s say you’re running a Performance Max campaign to drive ticket sales for a museum with great outdoors space for kids and you have limited resources to develop new landing pages that highlight this benefit. Since Performance Max uses your website as an important source to help find relevant search queries, you might have limited coverage and not show up on these valuable themes. That’s where search themes come in—you can add “activities for children” and “outdoor recreation” as search themes, to ensure you reach customers across channels that are interested in those offerings.

This concept is quite similar to the modern way of advertising on Meta, which has gone completely broad. You give it some creatives and an objective and it figures how to deliver that to you.

Except Google actually gives you multiple ways to train its algorithm and it’s in your best interest to spoon-feed it as many signals as you can to get the best results.


As an example, here are the most powerful inputs to get a furniture brand who sells modular couches better outputs from Google:

  • Thematic adgroups which are groupings of similar search keywords within an adgroup
    • Keywords consisting of: modular couch, interchangeable couch, easy to move couch, rearrange couch
  • Ad creative which should be related to the theme of the given adgroup
    • Headlines and descriptions that speak to modular couches and specifically mention the top searched terms for this product
  • Landing page content which is crawled by Google’s paid bot to decipher what you’re selling on your page (only when you use broad match keywords)
    • A dedicated landing page that features strong callouts and loads of helpful content about modular couches
  • Audience signals which highlights the types of audiences and users that are likely to convert so Google gets a better idea of the types of people it should target for you
    • Customers who have bought modular couches from you, an audience of users who have been to your modular couch PDPs, users who have been to competitors who also sell modular couches

Prior to this shift in the way that Google works, none of these were considerations in how Google chose which search queries to serve your ads in. It looked at your keyword, your bid, and your quality score to determine if your ad was relevant enough and your bid high enough to serve.

But now this game is all about content, specificity, and signals.

Similar to how marketers are starting to talk about how creative is the new targeting on Meta — it’s becoming quite clear how signals are the new targeting in Search.

So how can you ensure your account is structured in a way to feed signals to Google? Glad you asked.

Here is how I like to design Search accounts nowadays:

  • One non-brand search campaign with as many adgroups as categories you have in your site navigation. Each adgroup should contain a group of relevant and similar broad-match keywords, an RSA (ad) that has headlines/descriptions/images related to the given theme, and a landing page that is specific to the theme.
    • On a weekly basis you should be going through the Search Terms report for this campaign and excluding irrelevant keywords while also adding in relevant search terms as keywords that Google's AI has matched
  • One PMAX campaign with a merchant center feed (if eComm) with asset groups covering the same categories you have in your site navigation. These asset groups need rich assets, audience signals, and search themes (duh).
  • One branded search campaign with adgroups for broad match, phrase/exact match, and any other brand related terms (eg. brand name + product name queries).

If you want a second set of eyes on your current setup and some specific direction on where to go, feel free to grab an express audit from me and I'll show you the way via a recorded video walkthrough of your account.

Hi! I'm Ben

I’m a CMO (and former Googler) helping DTC brands and online retailers make sense of the things that matter. Subscribe to my newsletter for my unique perspectives, relevant data, and ways to grow your business.

Read more from Hi! I'm Ben

Over the past few months, I’ve consulted with a few eCommerce brands who have been struggling to scale PMAX up to levels they used to achieve with Smart Shopping. Each have had their own theories (and conspiracies) about the reasons why this is the case but none had taken the time to dig into what was going on with their campaign and underlying product catalogs. My first move in these cases is to run this PMAX script built by Mike Rhodes (not sponsored) which surfaces a number of insights...

It’s been nice to be on the brand side of Black Friday week for the first time in my career. Up until this year, my role during BFCM has always been trying to tell other brands what I’d suggest they do and hoping they did it (none of them did). And now I can do what I want :) Turns out that one of the key things I always talked about as a Google rep worked out for one of my stores… In the months leading up to November, I had always pushed the concept of lowering efficiency in exchange for...

With BFCM approaching I figured I’d drop in and provide my perspective on the number one thing people mess up with Google Ads on BFCM… Accounting for conversion lags and conversion windows in their reporting and strategy. Google Ads attributes conversions differently than many expect and it’s a far departure from how performance is reported by Google Analytics, Northbeam, and Facebook. This lack of understanding leads to an interpretation using incorrect data which leads to bad decision...