Google Search campaigns used to feel easier to define.
You chose the keywords. You selected exact, phrase, or broad match. You wrote the ads. You reviewed search terms and added negatives. The keyword structure created most of the boundaries around where the campaign could appear.
Those controls still exist, but Google now uses much more than the literal keyword when deciding whether an ad may be relevant.
Broad match can connect a keyword with searches that use completely different wording. Exact and phrase match can include searches with the same meaning or intent. AI Max can go further by learning from keywords, ads, and website content to find additional queries.
Google can also customise ad text and, when enabled, choose another landing page that it considers more relevant to the search.
That creates an understandable question:
If Google controls more of the matching, messaging, and landing-page selection, can advertisers still control Search campaigns?
Yes.
But control now means something different.
It is less about predicting every query before it happens and more about defining where Google may explore, what counts as a valuable conversion, which pages and brands are eligible, and where automation needs to stop.
Search control moved from keywords to guardrails
Keywords still matter.
They help describe the commercial themes a campaign wants to capture and remain an important part of account structure.
But Google no longer reads them only as strings of words.
Matching can also use information such as:
- The meaning and intent of the search
- Other keywords in the ad group
- Ad content
- Landing-page content
- Previous search activity and other contextual signals
- Conversion predictions
- The selected bidding strategy
With AI Max, eligible searches can also be found through asset- and landing-page-based matching without requiring an existing keyword to provide the direct starting point.
This gives Google more freedom.
It also changes the campaign manager's job.
We no longer ask only:
Which keyword should match this query?
We also need to ask:
Should this query be eligible at all? Which page should receive it? Which message may be shown? And does the conversion data give Google the right definition of success?
Exact, phrase, and broad still create different boundaries
The three match types still exist, but their names should not be read literally.
Exact match
Exact match gives the tightest match-type control.
But exact does not mean identical wording.
An exact-match keyword can match searches that Google considers to have the same meaning or intent, including eligible close variants.
That makes exact match useful when we want a focused theme around an important commercial search without requiring customers to type one exact sequence of words.
For example, a store may want particularly clear reporting and control around a high-value product category or a very specific buying intent.
Phrase match
Phrase match gives Google more reach while still requiring the search to include the meaning of the keyword.
It can be useful when we understand the central commercial theme but want to capture additional ways customers express that need.
Phrase match therefore sits between exact and broad in terms of reach.
It still needs search-term review. A search can be linguistically related to the keyword while having a completely different commercial purpose.
Broad match
Broad match gives Google the most flexibility.
Google can use the keyword together with additional contextual signals to find related queries, including queries that do not contain the original keyword terms.
Google recommends using broad match with Smart Bidding because bidding can evaluate the individual context of each auction rather than applying the same decision to every search matched with the keyword.
Broad match can uncover useful demand that the advertiser would never have added manually.
It can also expand into searches the business does not want.
The important point is:
Broad match is not simply a bigger keyword. It gives Google's matching system more room to explore.
Whether that exploration is useful depends on what surrounds the keyword.
Exact and phrase match still have a job
Google increasingly encourages broader matching with conversion-based Smart Bidding.
That does not mean every Search account should immediately replace all exact and phrase keywords.
Tighter match types can still be useful when:
- A specific query has unusual commercial importance
- The product has several easily confused meanings
- Regulatory or messaging constraints require tighter matching
- The account is being diagnosed and clearer query boundaries are useful
- Conversion measurement is still being validated
- A business deliberately wants a narrow Search campaign
Imagine a company selling specialised professional equipment.
Searches around the same product name might include:
- People wanting to buy the equipment
- Students looking for instructions
- Technicians looking for spare parts
- People looking for second-hand products
- Job seekers
- People looking for free downloads or manuals
Broad match may identify valuable patterns over time.
But the business may still prefer tighter initial boundaries around its most important commercial searches.
No match type is automatically correct for every account.
What happened to campaign-level broad match?
Google introduced a campaign-level broad match setting that converted existing exact and phrase keywords to broad match and made new keywords broad by default.
That setting is now being folded into Google's wider Search automation strategy.
Starting in September 2026, Google says campaigns using the campaign-level broad match setting are being automatically upgraded to AI Max.
This means we would no longer treat campaign-level broad match as the long-term strategic choice it once appeared to be.
The more relevant decision going forward is how AI Max is configured and which of its controls the campaign actually needs.
AI Max is not another keyword match type
This distinction is important.
AI Max for Search is a wider automation layer rather than a fourth match type.
Google currently groups it into two main areas:
- Search term matching
- Asset optimisation
Search term matching
AI Max search term matching can expand beyond existing keywords using broad-match and keywordless technologies.
Google can learn from:
- Existing keywords
- Ads and other creative assets
- Landing pages
That creates a significant difference from ordinary broad match.
A broad keyword still starts with a keyword.
AI Max can also use the meaning of the ad and the website to identify additional relevant searches.
Asset optimisation
The other side of AI Max affects how the ad and destination can change.
This includes:
- Text customisation, which can create or adapt ad text
- Final URL expansion, which can select a different relevant landing page
The result is that Search automation can now affect three different stages:
Query → message → landing page
That is why AI Max needs to be reviewed as more than another targeting checkbox.
Broader matching makes conversion quality more important
Giving Google more room to find queries only works if Google is learning from a useful business outcome.
Suppose one account optimises towards:
- Real completed purchases
- Correct transaction values
- Reliable customer and product data
Another optimises towards:
- Duplicated Purchase events
- Add-to-cart actions counted as primary conversions
- Incorrect revenue values
- Several weak conversion actions mixed together
Broad matching increases the number of opportunities available to both campaigns.
But they are not learning the same lesson.
The first campaign is being taught which searches create valuable orders.
The second may simply become better at generating the wrong signals.
Automation does not decide whether your conversion goal is commercially meaningful. It assumes you chose correctly.
This is why conversion tracking becomes more important as matching becomes broader.
We cover that implementation in more detail in our guide to Google Ads conversion tracking for eCommerce.
Negative keywords still matter
Automation did not remove one of the most useful Search controls: telling Google what you do not want.
Negative keywords can prevent ads from appearing on queries that match the excluded terms according to the selected negative match type.
For eCommerce accounts, they can be useful for clearly irrelevant intent such as:
- Products the store does not sell
- Jobs or careers
- Free downloads
- Repair services
- Instructions or manuals where they have no commercial value
- Second-hand products
- Locations the business cannot serve
- Incompatible products or models
- Terms that create a brand-safety problem
But negative keywords are restrictive.
Google currently recommends using them mainly for clearly irrelevant traffic and important brand-safety exclusions rather than trying to sculpt every uncertain query manually.
That recommendation makes sense.
The job of a negative keyword is not:
This search has not converted yet.
It is closer to:
This search conflicts with what we sell, who we serve, or how the business should appear.
Do not judge a search term too quickly
One non-converting click is not evidence that a query is bad.
Neither are five clicks in many accounts.
The amount of evidence required depends on:
- Typical conversion rate
- Average order value
- Product consideration period
- Cost per click
- Margin
- Available conversion volume
A clearly irrelevant query can be excluded immediately.
A commercially plausible query may need more data.
This distinction prevents two opposite mistakes:
letting obvious waste continue and removing useful new demand before it has a chance to prove itself.
Negative match types do not behave like positive match types
Negative broad, phrase, and exact match have their own matching rules.
They also do not automatically expand to every related synonym, singular, plural, or semantic alternative in the same way Google's positive keyword matching can understand meaning.
That matters when building exclusions.
A negative keyword list should therefore be reviewed for both:
- Traffic it successfully removes
- Useful searches it may accidentally block
A large negative list is not automatically a well-managed list.
Good negative keyword management removes irrelevant spend without unnecessarily closing useful routes to demand.
Use brand controls when the issue is a brand
Brands deserve separate treatment.
Google now recommends using dedicated brand controls instead of ordinary negative keywords when the advertiser specifically wants to include or exclude branded searches.
Brand exclusions can prevent a campaign from serving on selected brand queries and account for related variants more comprehensively than manually creating a list of brand negatives.
AI Max also supports brand inclusions and exclusions.
This can help when an advertiser wants to separate:
- Its own brand
- Competitor brands
- Manufacturers
- Retailers or marketplaces
- Specific branded product ranges
Brand traffic is not inherently a problem.
It simply has different economics and intent from non-brand acquisition.
If the two are blended together, a strong brand result can make broader prospecting performance look healthier than it actually is.
Text customisation gives Google more control over the message
AI Max text customisation can generate or adapt headlines and descriptions using information from the campaign, existing ads, landing pages, and Google's generative systems.
This can help when the campaign reaches a query that is valuable but more specific than the manually written ad.
The advantage is obvious: Google can create more query-specific combinations at a scale an advertiser could not write manually.
The trade-off is equally important:
the advertiser gives up some direct control over the final wording.
That matters when:
- Product claims need approval
- Pricing has conditions
- Delivery is available only in certain situations
- A regulated category requires exact wording
- The brand deliberately avoids certain claims or phrases
Google itself advises advertisers to monitor generated assets for accuracy.
The quality of automated copy also depends on the quality of the source material.
If landing pages are vague, outdated, or poorly structured, Google has weaker material to work from.
Final URL expansion turns the website into a targeting input
With Final URL expansion, Google can select another relevant page from the advertiser's site instead of always using the manually provided landing page.
This can work well for a large eCommerce site.
A search for a very specific product type might be sent directly to the relevant subcategory instead of a broad category page.
But semantic relevance is not the same as commercial quality.
A website may contain pages that are related to the search but should not receive advertising traffic.
Examples include:
- Old campaign pages
- Out-of-stock product ranges
- Weak categories
- Informational articles
- Support pages
- Low-margin products
- Pages with poor mobile performance
This is why AI Max also provides URL controls.
Campaign-level URL exclusions can prevent selected pages from being used.
Ad-group-level URL inclusions can give the system more specific guidance about which parts of the site are relevant.
The website is no longer only where the ad sends traffic.
Its structure and content also help determine which queries Google believes the campaign can answer.
Pinned assets need special attention with URL expansion
Advertisers sometimes pin responsive search ad assets because particular wording must appear in a specific position.
That control interacts with AI Max.
Google currently states that when Final URL expansion selects another landing page, pinned RSA assets may not be respected because the manually pinned message may no longer match the selected page.
URL inclusions can create a similar issue.
If fixed messaging is a business requirement, that needs to be considered before enabling those features.
In some campaigns, tighter URL control or disabling Final URL expansion may be more important than the additional reach.
AI Max still gives us search-term visibility
More automation does not mean Search has become a complete black box.
Google currently provides AI Max reporting across several areas, including:
- Search terms
- Keywords
- Ads and assets
- Landing pages
The search-terms report can identify AI Max traffic and show whether expansion came from broad-match or keywordless matching.
Google also provides a view that combines the search term, headline, and landing page.
This is particularly useful because those three things now need to be evaluated together.
A search term may be relevant while the selected page is poor.
The landing page may be strong while the generated headline makes an inaccurate promise.
Or all three may work together well and reveal a commercial theme worth developing further.
Human review now needs to look at the whole search journey
When we review an automated Search campaign, we do not want a report containing only keyword performance.
We want to understand:
| Question | Why it matters |
|---|---|
| Which queries are expanding beyond the keyword list? | Shows where automation is discovering new demand |
| Which queries clearly do not fit? | Identifies negative-keyword opportunities |
| Which landing pages are being selected? | Checks whether traffic reaches commercially useful pages |
| What ad text is being generated? | Checks message accuracy and brand fit |
| How much performance comes from brand? | Adds context to acquisition performance |
| Which products and customers are converting? | Connects platform performance to business value |
| Which new themes deserve their own structure? | Turns useful automated discovery into better campaign architecture |
This is not less campaign management.
It is campaign management at a different layer.
Where we would give Google less freedom
Broad match and AI Max are not automatically wrong for a specialised or smaller account.
But some conditions justify more caution.
We would be particularly careful when:
- Purchase tracking is unreliable
- Primary conversion goals do not represent the real business outcome
- The account has very little useful conversion history
- The product has several unrelated meanings
- The website mixes commercial and informational content heavily
- Advertising claims need strict approval
- Stock changes frequently
- Different markets have very different economics
- Brand and non-brand results must remain separate
- The budget leaves little room to test weak traffic
That does not mean automation must remain disabled forever.
It means the account should first improve the inputs and boundaries that automation depends on.
Do not confuse additional reach with profitable growth
AI Max and broad match can find queries an advertiser would not have predicted.
That is valuable only if those queries generate customers the business actually wants.
A Search conversion can still be commercially weak.
It may:
- Sell a very low-margin product
- Come almost entirely from existing branded demand
- Generate a customer likely to cancel or return the order
- Use a discount that removes most of the contribution
- Send traffic to a category the business does not want to scale
This is the same principle we apply throughout automated advertising:
Google can optimise the outcome it can see. The business still has to decide whether that outcome is worth scaling.
A practical control framework for modern Search
Instead of asking whether Search is “manual” or “automated,” we prefer to map the controls that remain available.
| Control | What it defines |
|---|---|
| Match type | How much matching flexibility begins with the keyword |
| Conversion goals and values | What Google's bidding systems treat as success |
| Smart Bidding | How bids respond to predicted conversion opportunity |
| Negative keywords | Which clearly irrelevant search intent must be excluded |
| Brand controls | Which brands should or should not participate |
| URL inclusions and exclusions | Which parts of the site automation may use |
| Text customisation settings | How much freedom Google has to generate messaging |
| Campaign structure | Which business objectives, markets, or products should remain separate |
| Search-term and landing-page review | Where automation is actually expanding |
| Business reporting | Whether conversions are commercially valuable after the click |
That is modern Search control.
Not deciding every query manually.
Deciding which freedoms Google receives and which ones it does not.
What we would review before enabling AI Max
Before turning on additional automation, we would check the surrounding system first.
1. Conversion tracking
Are primary conversions real business outcomes? Are purchases deduplicated? Are values and currencies correct?
2. Search terms
Does the existing account already show obvious irrelevant themes that should be excluded?
3. Website structure
Are commercial categories clear? Are there outdated, informational, unavailable, or otherwise unsuitable pages that should not receive paid traffic?
4. Ad messaging
Can the landing pages support generated copy accurately? Are there claims that must never be generated automatically?
5. Brand strategy
Should own-brand and competitor-brand traffic be included, excluded, or separated?
6. Commercial economics
Are all products equally valuable to advertise? Are there low-margin categories, stock constraints, or markets that require separate treatment?
Once those answers are clear, AI Max becomes much easier to evaluate.
Test expansion instead of arguing about it
There is no reason to treat AI Max as a philosophical choice between “trust Google” and “never give up control.”
Google currently supports AI Max experiments for existing Search campaigns.
That makes a controlled test a much better option where enough traffic and conversion volume are available.
Compare the result against the existing campaign and look beyond total conversion count.
Review:
- Incremental queries
- Landing pages
- Generated assets
- Conversion value
- Customer and product quality
- Brand contribution
- Actual backend sales
The question is not whether AI Max generated more traffic.
It is whether the additional freedom created a better business result.
Control is not gone. It is distributed.
The old Search account put much of its control inside the keyword list.
The current Search account spreads control across keywords, bidding, conversion data, negatives, brands, assets, URLs, reporting, and the website itself.
That can feel less predictable because advertisers no longer approve every possible query-and-ad combination in advance.
But it does not mean Search campaigns have become uncontrollable.
It means control has moved from micromanaging every match to designing the system in which matching happens.
At Creative Brackets, we treat Search as a connected query, ad, landing-page, and conversion system rather than a keyword list.
That means reviewing what Google is discovering while protecting the boundaries the business cannot afford to automate.
Not sure whether automation is finding new demand or simply expanding your spend?
FAQ
1. Does exact match still mean exact?
No. Exact match can show ads on searches that Google considers to have the same meaning or intent as the keyword, including eligible close variants. It remains the narrowest of the three positive keyword match types, but it should not be interpreted as literal word-for-word matching.
2. Is broad match better than phrase or exact match?
Not universally. Broad match provides the widest reach and is designed to work with Smart Bidding. Exact and phrase can still be useful when tighter matching is important. The right choice depends on the objective, measurement quality, product, budget, and level of control the business needs.
3. What is AI Max for Search?
AI Max is a set of Search automation features rather than a keyword match type. Its search term matching can use broad match and keywordless technology to find additional queries. Its asset-optimisation features include text customisation and Final URL expansion.
4. Are negative keywords still useful with AI Max?
Yes. They remain a direct way to block clearly irrelevant or unsafe search intent. Google recommends using them carefully because they can restrict useful traffic. When the goal is specifically to exclude a brand, Google recommends using brand exclusions instead.
5. Can AI Max send traffic to a different landing page?
Yes, when Final URL expansion is enabled. Google can choose another relevant URL from the website. Advertisers can use URL controls or disable the feature when tighter landing-page control is required.
6. Can Google change my Search ad copy?
With AI Max text customisation enabled, Google can generate or adapt text assets using information from your ads, landing pages, and other campaign data. Those generated assets should still be reviewed for accuracy and compliance with the brand's requirements.
7. Does AI Max hide search terms?
No. Google provides AI Max reporting that includes search terms, keywords, assets, and landing pages. Its dedicated search-term views can also show the search term together with the headline and landing page used.
8. Should I enable AI Max on every Search campaign?
No setting should be enabled only because it is available. AI Max makes more sense when conversion measurement is reliable, the website contains suitable landing pages, the campaign has clear commercial goals, and the advertiser has defined negatives, brand controls, and other important boundaries. Where possible, testing it through an experiment can provide better evidence than applying it across the account immediately.