Keyword research in 2026: still relevant in the AI era?

Keyword research is still relevant in 2026, but the way you do it has fundamentally changed because of AI. Keyword research is the process of identifying the search terms your target audience uses, so your campaigns can respond to the right intent. Anyone who thinks AI has fully taken over this work is missing half the picture.

Key takeaways

• Keyword research remains the strategic foundation of every Google Ads campaign, even in the AI era.
• AI tools like Smart Bidding and Performance Max process search intent automatically, but need quality input to work well.
• Broad match and Performance Max are powerful, but dangerous without a tight system for negative keywords and search term monitoring.
• The AdBrains Keyword Incubator safely tests new keywords before they enter production, limiting budget waste.
• Intent analysis replaces simple search volume hunting: the value of a keyword lies in what the searcher wants to do, not just how many people type it.

What has actually changed about keyword research?

Until a few years ago, keyword research was a manual, periodic task: open Google Keyword Planner, check volumes and competition, build a list, and load it into your campaign. In 2026, that model is outdated. Not because keywords no longer matter, but because the dynamics have fundamentally shifted.

Google processes billions of searches daily, of which approximately 15% are entirely new, according to Google Search Central (2023). That means they have never been entered before. Voice search, AI-generated queries and conversational search have dramatically increased the diversity of what people type. A static keyword list is structurally too slow to keep up with what people are actually searching for.

At the same time, Google has shifted its match type logic. Broad match in 2026 captures a much wider range of variants than before, and Performance Max decides for itself which queries are relevant based on your assets and signals. That sounds convenient, but it cuts both ways: more reach also means more risk of irrelevant impressions and wasted budget.

Search intent is the new search volume

Search intent distribution in typical Google Ads campaigns

Example: distribution by intent type in search campaigns

Commercial intent (buy, quote)

38%

Informational intent (how, what, why)

29%

Navigational intent (brand name, URL)

18%

Comparative intent (best, review)

15%

The shift from volume-based thinking to intent-based thinking is the biggest strategic change for anyone doing keyword research. Targeting a keyword with 10,000 monthly searches makes little sense if three-quarters of those searchers are still in the orientation phase and have no intention of buying or requesting a quote anytime soon.

Take ToetsJeKennis.nl, a platform for online exams and courses with an average order value of around €50. The gap between someone searching for "practice exam free" and someone searching for "buy exam training" is enormous. One reflects informational intent, the other commercial intent. Advertisers who fail to separate those two in their keyword structure end up paying for clicks that go nowhere.

Lead generation clients face exactly the same problem. Clima-Active.nl, an installer of air conditioning and heat pumps, is a good example. Queries like "heat pump disadvantages" or "how does a heat pump work" are informational. Someone typing "heat pump installer quote" or "air conditioning installation price" is ready to act. In 2026, solid keyword research begins with intent segmentation rather than chasing the highest search volumes.

Broad match and Performance Max: power with risk

Broad match has never been more capable than it is in 2026, but running it without close monitoring is asking for trouble. Google's own data (Google Ads Help, 2024) shows that broad match paired with Smart Bidding delivers more conversions on average than phrase or exact match alone. That advantage evaporates the moment irrelevant search terms start eating through the budget before Smart Bidding has had any real chance to learn.

Performance Max has shifted what keyword research even means in practice. There are no explicit keywords to specify; you supply assets, audiences and URL signals, and Google's algorithm takes it from there, deciding which queries your campaign appears for. Keyword research becomes less visible in the interface as a result, though it matters just as much as before. Pmax search themes offer one of the few remaining direct levers, letting you nudge the algorithm toward specific query clusters without reverting to a traditional keyword list. What you put into your asset texts, landing pages and audience signals is what determines how accurately Performance Max matches your ads to actual intent.

For E-4motion.com, the webshop for new electric folding bikes, this plays out in a straightforward way: vague product texts and ad assets pull in a wide audience that rarely converts. Terms with real purchase intent behind them, like "buy electric folding bike", "lightweight foldable e-bike" and "electric bike for the train", point the algorithm somewhere worth going.

How AdBrains automates keyword research with AI

Keyword management: manual vs. AI-driven

What changes when AI takes over keyword strategy?

Manual management

• Periodic keyword research (monthly or quarterly)
• Limited visibility into new search terms
• Negative keywords added manually after analysis
• Broad match risky without continuous monitoring
• Campaign expansion is time-consuming

AdBrains AI managementAI advantage

• Daily automated search term mining
• Automatic detection of irrelevant search terms
• Negative keywords added immediately without delay
• Broad match safely used with AI monitoring
• Automatic campaign expansion for growing search terms

At AdBrains, we haven't replaced keyword management with AI. We've turned it from a periodic task into a continuous, automated process, and that distinction matters far more than it might sound when you're thinking about campaign quality over months and years.

Our automated search term mining system goes through every search term that generates actual traffic in active campaigns, every single day. Based on conversion performance, semantic analysis and pre-set exclusion rules per client, the system flags irrelevant terms and adds them as negative keywords straight away, without queuing them up for a weekly or monthly review. For a client like Clima-Active.nl, where a click on the wrong term means money gone with nothing to show for it, that immediacy has a measurable effect on the cost per quote request.

Newly detected terms that look potentially relevant don't go straight into the production campaign. They land in our Keyword Incubator first: a separate campaign environment where new keywords run on a limited budget until they've accumulated enough conversion data to be judged fairly. Only once a keyword shows a positive return does it get promoted automatically. For a client like ToetsJeKennis.nl, this keeps unproven keywords from quietly eroding the campaign budget during those early weeks when they could just as easily flop as perform.

On top of that, our multi-agent verification system reviews every optimisation decision before anything is actually executed, whether that's adding a negative keyword, promoting a term or adjusting a bid. Four independent AI agents check each action. This catches the kind of errors that tend to slip through in fully autonomous setups, and it means our clients can trust that every change is there for a reason.

Our auto campaign expansion system watches for search terms that are gaining momentum. When a cluster of related terms consistently builds volume and conversions, the system picks up the signal and expands the campaign structure with new ad groups or campaigns automatically. Manual keyword research simply cannot respond at this pace. Where a human review cycle might catch a trend after a month, the system acts within a day.

Finally, our server-side signal enrichment strengthens the conversion data that Smart Bidding relies on. Through our own sGTM infrastructure, we connect first-party client data to conversion signals, so Google's algorithm understands not just that a conversion happened, but what it was worth. For clients with several product categories, this feeds into a more accurate tCPA or tROAS strategy at the product group level.

The Keyword Incubator in practice

Keyword Incubator: from new keyword to production

Example: throughput per phase for a new campaign

New keywords detected

All candidates from search term mining

Added to incubator campaign

After filtering on relevance and intent

Sufficient conversion data collected

After test period in incubator

Promoted to production campaign

Proven keywords with positive return

28% of candidates reach production

The Keyword Incubator is one of the most concrete implementations of how AdBrains has transformed keyword research. In the traditional model, new keywords are added directly to an existing campaign, where they compete for budget against already proven performers. The result is that unproven keywords drag down the campaign average before you have enough data to judge whether they are worth it.

In the Incubator model, candidate keywords are first tested in isolation. They receive their own impressions, their own clicks and their own conversion data, without clouding the statistics of the main campaign. The thresholds for promotion to production are configurable per client: an e-commerce client with a short purchase funnel uses different minimum conversions than a lead generation client where each quote request represents higher value.

As an example: suppose a campaign for HACCP-cursus.com receives ten new keyword candidates after a search term mining cycle. Those ten are placed in the incubator with a limited test budget. After two weeks, three keywords have sufficient conversion data with an acceptable CPA, four perform ambiguously, and three have zero conversions at significant click volume. The three proven keywords move automatically to production, the four ambiguous ones get an extended test period, and the three non-performers are automatically paused. No manual decisions, no delayed reviews.

Keyword research step by step: what still works in 2026?

Automation has taken over a lot of the grunt work, but the strategic layer still needs a human hand. Below are the steps that actually move the needle in 2026.

1. Start with intent segmentation: Split your keyword universe into commercial, informational, navigational and comparative intent. Conversion campaigns belong in the commercial and comparative bucket; informational intent is better served by content campaigns or display activity.
2. Use Google Search Console as your primary data source: Search Console shows which queries are driving organic traffic to your site, including queries you haven't started bidding on yet. It's an underrated place to find new keyword candidates, and most advertisers ignore it.
3. Analyse search term reports daily (or automate it): Broad match and Performance Max keep generating new match variants, often faster than you'd expect. Skip daily monitoring and you'll quickly lose sight of where the budget is actually going.
4. Build a tight negative keyword policy: Maintain exclusion lists at both campaign and account level, and revisit them at least once a week. Competitor brand terms, free-related terms for paid products and irrelevant synonyms are the obvious starting points.
5. Test new keywords in isolation: Run new candidates through an incubator setup so they don't drag down the performance of keywords that are already working.

Keyword research vs. Performance Max

The takeaway from this comparison: even with Performance Max, keyword thinking is relevant. You simply translate it differently: into the choice of asset texts, the audience signals you provide, and the negative keywords you set at account level to make clear what you do not want to reach.

What you should never leave entirely to AI

• Brand strategy: The decision whether to advertise on your own brand name, that of competitors or partners involves legal, strategic and financial dimensions that require human judgement.
• Seasonal priorities: Clima-Active.nl has a different campaign focus in summer (air conditioning) than in winter (heat pumps). An AI without business context will not independently initiate that strategic shift.
• New product launches: When E-4motion.com introduces a new electric folding bike model, there is no historical search term data yet. Initial keyword research based on market, competitor and audience analysis is purely human work at that point.
• Budget allocation across campaign types: The strategic choice of how much budget goes to Search, Performance Max or other campaign types remains a human decision based on business goals and market position.

Frequently asked questions about keyword research in 2026

Is keyword research still necessary if I only use Performance Max?

Yes, absolutely. Performance Max does not require direct keyword input, but the quality of your campaign depends directly on how well you steer the algorithms. That steering comes through your asset texts, your audience signals and your account-level negative keywords. Without keyword research, you cannot formulate those inputs well and you lose control over what audience the campaign attracts. Keyword research with Performance Max is indirect, but no less essential.

What is the difference between a keyword and a search term?

A keyword is what you enter in your Google Ads campaign as a trigger for your ad. A search term is what the user actually types into Google. Through broad match and Performance Max, one keyword can activate dozens or even hundreds of different search terms. Search term mining is about analysing those actual search terms, not just the keywords you have specified. This distinction is fundamental to understanding where your budget goes.

How often should I review my keyword strategy?

In the traditional model, you did this monthly or quarterly. In 2026, a daily automated system for search term monitoring is the standard. The strategic level, covering intent segmentation, campaign structure and seasonal adjustments, should be reviewed at least quarterly, or when there are significant market shifts. Operational decisions such as adding negative keywords and promoting incubator keywords need to happen daily or weekly to prevent budget waste.

What makes a keyword "good" in 2026?

A good keyword combines three properties: relevant intent (the searcher wants what you offer), sufficient volume (enough searches to generate data) and realistic competition (the CPC fits within your Target CPA or Target ROAS). Search volume alone is no longer a good indicator. A keyword with 200 monthly searches and high commercial intent can be more valuable than a term with 10,000 searches and mainly informational intent, especially in e-commerce with a specific audience like HACCP-cursus.com.