Search Isn’t Matching Intent — It’s Matching Predictions: Inside the Predictive Era
TL;DR: The Death of Keyword Precision
The Big Idea: Google Ads has fundamentally pivoted from a Literal Era (matching your keywords to queries) to a Predictive Era (matching users to outcomes). In 2026, the search query is no longer the command; it is merely one of thousands of signals the AI uses to predict whether a specific user is worth your bid.
I have been sitting on this opinion piece for a while.
As a PPC consultant and a strategist, I am cautious about "twisting words" or misrepresenting the platforms we rely on.
However, after spending 17 years working in Google Ads and PPC and observing account behavior as a coach. I feel ready to say my official thoughts regarding where Google Ads and frankly other ad platforms are heading for the future.
Google Ads is no longer matching keywords on intent but rather they are a predictive platform.
My point of view was born from the delta between what Google says and what the data does in the accounts that I am advising on strategy.
The Great Drift: From Literal to Predictive Matching
For years, the PPC industry has obsessed over match types and campaign structures.
It’s still the dominant agency pitch — “The Best Google Ads Setup for 2026” or “Copy This Strategy to Explode Your Results.”
It is a “we do it better” way of messaging to get clients.
But prescriptive setups and a “better way” don’t fix what struggling businesses actually face.
While the industry makes claims about phrase vs. exact or PMax vs. Shopping, the real issue is that the entire system has drifted and is failing many businesses.
I see advertisers with "perfect" setups failing all the time. The reason? They are still operating under the Old Model.
The Old Model (The Literal Era): For most of search marketing’s history, the logic was linear. The query was the command.
Query → Keyword → Ad
Even when match types loosened, the underlying logic stayed the same: a user expressed intent through a query, Google matched that query to your keyword (first literally, then with intent interpretation), and your ad appeared.
A user typed a query
Google matched it to your keyword (literal then intent based over the years)
Your ad showed
The New Model (The Predictive Era):
In 2026, Google has fully shifted to a logic that prioritizes predicted outcomes over literal requests. I’ve been arguing this for the past two or three years, but I didn’t yet have the volume of data — or the language and thinking models — to articulate the case as clearly as I can now.
(User Signals + Journey History) x Creative Meaning = Predicted Conversion
Instead of matching a query to a keyword, Google evaluates a blend of behavioral signals, past interactions, contextual cues, and the semantic meaning of your creative. The model then chooses whatever combination of keyword, campaign, and asset it believes will produce the highest probability of conversion.
It’s no longer “What did the user type?” It’s “What is this user most likely to do next?”
Prediction, not precision, drives the system now.
If you would prefer to watch me talk about this concept
Evidence 1: Accuracy is No Longer the Priority
In the "Old Model," negatives were a hard wall.
In 2026, they feel like suggestions, during #PPCChat Ginny Marvin confirmed that for new surfaces like AI Overviews and AI Mode, ads are matched to a "deeper intent" and require Broad Match or keywordless targeting.
Even more revealing was during the #PPCChat the response to why Google substitutes product model numbers in ads.
The official answer? It happens if a product is "not predicted to result in a conversion."
Let that sink in: Google will ignore a user’s specific request for "Model A" and show them "Model B" if the algorithm thinks "Model B is more likely to get a conversion. This isn't a "better match" this is the system prioritizing liquidity and outcomes over literal accuracy.
Evidence 2: The "Journey" is the New Keyword
With the upcoming rollout of Journey Aware Bidding (JAB), the search query is no longer the primary signal. JAB documentation says, it allows "non-biddable" journey stages (like a site visit or a middle-of-funnel download) to influence the bid for a primary conversion.
You aren't paying for just keyword anymore.
You are in an auction based on Google's ability to predict a user's behavior based on signals you can't even see.
The "Prediction" isn't just about the word; it’s about whether a user in this specific context that you allow and feed the platform plus signals you can’t see (battery life, device price point, browsing history things we don’t know because of privacy) is worth the auction price.
Evidence 3: The Liquidity Mandate
The introduction of Campaign Total Budgets (CTB) for Search and Pmax reinforces this. CTB removes daily spending limits to ensure your budget is fully utilized by an end date. Google even recommends "relaxing targets" to ensure the money is spent.
When you combine Broad Match requirements in AI surfaces with Total Budgeting, the system stops being a filter for your traffic and starts being an Accelerator for AI to learn and predict.
Evidence 4: Google’s "Smart Matching"
I don’t know why I didn’t connect this sooner with “close variants.” If the AI predicts a conversion, it will often find a way around a negative keyword — either by treating the query as a close variant or by reclassifying it under an intent cluster that isn’t blocked. I’m basing this on years of watching real account behavior, not theory.
This is why we see competitor brands matched as “close variants”.
The Shift from Query to Identity
What is often lost in this transition is that Google is no longer bidding on the query; they are bidding on the user.
In the Old Model, a keyword was a universal bucket. In 2026, the auction price for the query 'best running shoes' is no longer the same for everyone.
Google is looking at the 'user model'—who is this person, what is their purchase history, and what is their 'value' to the ecosystem?
You aren't competing for a spot on a search results page; you are competing for a specific person's attention based on a profile Google has built behind the scenes.
| Feature | The Old Model (Literal) | The New Model (Predictive) |
|---|---|---|
| Primary Signal | The Search Query (Keywords) | The User Profile (Identity/Behavior) |
| Bidding Logic | How much is this word worth? | How much is this person worth? |
| Match Precision | Literal or Semantic Accuracy | Predicted Probability of Conversion |
| Role of Negatives | A hard wall/exclusion | A directional signal/suggestion |
| CPC Driver | Competitor bids on the keyword | Predicted value of the specific user |
This is why 'intent' has taken a backseat to 'prediction.
This shift is the primary driver of CPC Inflation. Because the system is no longer bidding on the 'keyword' but on the 'predicted outcome,' it justifies an aggressive bid based on the user's perceived value.
You might see a $15 CPC for a keyword that historically cost $5. This isn't because competition increased but because Google 'predicted' this specific user was a high-value target and charged you a higher CPC to reach them.
Why This Matters: The "Minimum Effective Dose"
Because prediction is a volume game, small and mid-sized businesses are facing a structural disadvantage from the very start.
There is a "minimum effective dose" of data which is roughly 30–50 conversions a month to have the slightest chance of success in the ad platform.
If you fall below this, the predictive engine enters "Exploration Mode" indefinitely.
You end up paying for a "learning phase" without ever reaching the "optimization phase."
This dynamic has always existed, even back when we were making decisions by hand. So this isn’t a place to blame the AI. What’s changed is that prediction modeling amplifies the gap. When everyone is bidding with predictive systems, you simply can’t compete using manual data, gut checks, and human‑built rules. Prediction scales; human judgment doesn’t.
Read my post: The Rule of 100 Clicks: Why Your Google Ads Team Must Resist Impatience
You aren't just fighting competitors; you are fighting the inherent margin of error in an AI that prioritizes probability over relevance.
This leads the bidding to occasionally 'hallucinate' a connection between a query and an ad, which is essentially an aggressive semantic expansion. This bid is just to see if it might convert. Even if it's a 'miss,' you still pay the higher CPC for that experiment.
The New Power Dynamic: Creative is the New Keyword
In this predictive model, Google trusts its own inputs more than yours. We see this in:
Creative Weighting: Google’s AI scans images and video to determine audience. If your image shows a luxury kitchen, Google will find "luxury" seekers even if your keywords are generic.
The Push of Value Based Bidding: By moving toward Value-Based Bidding (VBB), Google can charge more for a "High Value" predicted conversion regardless of keyword competition. You aren't paying for the conversion; you’re paying a higher bid on the predicted conversion.
Intentional Opacity: From limited PMax reporting to "vanishing" Demand Gen conversions, the system is becoming a black box. This opacity isn't just a bug, it protects the predictive model and how it works because we would want to make changes.
How to Advertise When Predictions Rule the Auction
If you made it this far down in the post then you have reached the part that I think makes me fundamentally different than most in the industry right now.
I ended up with the POV I have because frankly I felt my “how to content” wasn’t allowing businesses to be fully successful.
But I found in my coaching work that once my clients understand that Google is optimizing for outcomes, not accuracy, they stop fighting the wrong battles.
Clients coaching with me have stopped obsessing over match types and trying to "force" the system into literal behavior.
Instead, they begin to focus on the only things that still move the needle in a predictive world:
Signal Strength: Feeding the AI better conversion data (offline conversions, enhanced conversions).
My warning: If you feed the predictive engine 'junk' data such as spam leads or low-value data, it creates a Feedback Loop of Failure, where the AI predicts and finds more of exactly what you don't want.
Read my full post on high value signals to get a better understanding of this concept: What Are High-Value Signals in Google Ads?
Creative Meaning: Ensuring your assets (images/video) clearly define your target audience for the AI's scanner.
Economic Alignment: Setting targets that allow for "Exploration Mode" without breaking the bank.
Landing Page Clarity: Making sure the "semantic meaning" of your destination matches the prediction.
This is the new skill set. You can’t steer the ship with a rudder anymore; you have to influence the current. Predictions are the decision-maker; advertiser inputs are now just suggestions.
Key Takeaways for 2026:
Predictions > Precision: Google will ignore exact match requests and negative keywords if its model predicts a high probability of conversion.
The User is the Target, Not the Word: Bidding has shifted from "what is this word worth" to "what is this specific person worth based on their identity and journey history."
The "Success Tax": Rising CPCs are often the result of Google predicting a "high-value" user and bidding aggressively on your behalf, regardless of keyword competition.
The Minimum Effective Dose: If your account generates fewer than 30–50 conversions per month, you are stuck in an indefinite "Exploration Mode," paying for the AI to learn rather than optimize.
