Keyword Research with AI: How to Find Profitable Keywords in 2026
Keyword research has changed dramatically over the last 18 months. Today, when I audit SMB projects in Catalonia, the first thing I check is whether they have adapted the process to the new capabilities of artificial intelligence. Most still work like it's 2022 —exporting massive lists from traditional tools— and are leaving value on the table. This guide explains the workflow we apply in 2026 with SMBs in Barcelona, Girona and Tarragona to identify keywords that are actually profitable, without falling for generic results or inflated data.
1. What changes in 2026: AI redefines keyword research
Until recently, keyword research meant throwing the main keyword into a tool and exporting 5,000 variants. In 2026, with AI Overviews, conversational assistants and the resurgence of natural-language searches, that method produces noise. AI, used well, lets us:
- Generate relevant semantic clusters in minutes (not hours).
- Detect the real intent behind a keyword without manual checking.
- Find very specific long-tail terms that don't appear in classic tools.
- Anticipate how a user might reformulate their search when talking to an LLM.
2. Which AI tools are worth using
Not every tool has the same strength. What I recommend —and what I use— is combining 2 or 3:
| Tool | Strength | Best for |
|---|---|---|
| ChatGPT / Claude / Gemini | Cluster and intent generation | Brainstorming and semantic grouping |
| Perplexity | Citation detection | Seeing which sites win AI Overviews |
| Ahrefs / Semrush | Real volume and difficulty | Validating economic potential |
| Search Console | Real data from your domain | Finding underused opportunities |
The mistake I often see in Catalonia is going all-in on a generative AI tool and ignoring real data: classic tools remain essential to validate that a keyword has enough traffic and a reasonable competitive context.
3. My step-by-step workflow with AI
This is the concrete process we apply on every new project. Total time has dropped from 6-8 hours to 2-3, while keeping strategic quality.
Step 1 · Brief the AI model with full context
Never give just the generic keyword. Provide context: sector, location (for example, a dental clinic in Tarragona), ideal client type, premium products and services, seasonality. The more context, the less generic the results.
Step 2 · Ask for semantic clusters, not lists
Instead of '100 keywords', ask for '10 thematic clusters of 8-12 keywords each, grouped by search intent'. The quality difference is night and day, and it lets you think in content, not loose words.
Step 3 · Validate volume with a classic tool
Load the keywords into Ahrefs, Semrush or Google Ads' Keyword Planner. Filter those with volume >50/month (in Catalonia, the real threshold is often lower; adjust to your market).
Step 4 · Ask the AI to infer dominant intent
For each keyword, ask whether it's transactional, informational, commercial or navigational, and what format the user would expect (article, guide, calculator, video). Verify by running the actual search on Google in incognito.
Step 5 · Identify AI Overview opportunities
Ask the model which keywords are candidates to have AI Overview in 2026 and which aren't. Then check against a real incognito search, because the AI can be wrong.
Step 6 · Prioritize by economic impact
Not every keyword with volume is profitable. Apply a volume × difficulty × conversion-value matrix. The AI can estimate conversion value if you give it LTV (lifetime value) and average margin for your business.
4. Common mistakes when doing keyword research with AI
The ones I most often find in audits of Catalan SMBs:
- Trusting volumes invented by the AI. LLMs don't know real volume; you must validate.
- Generating a thousand keywords without grouping by intent (result: messy content).
- Ignoring conversational long-tail ('how can I...', 'is it worth...').
- Not separating keywords with AI Overview from those without (content strategy differs).
- Skipping Search Console validation for your own domain.
5. How to validate that a keyword is actually profitable
A keyword with volume is not the same as a profitable keyword. To validate:
- Check the indicative CPC in Google Ads: if it pays >0.50 €, it has real commercial intent.
- Look at the top 10: if it's dominated by giants (Amazon, Wikipedia, big media), abandon or find a local angle.
- Calculate lead value: traffic × CTR × conversion × average customer value.
- Check for AI Overview: if there is one, plan citable content rather than chasing mass traffic.
6. Combining AI with real data: the best of both worlds
What I always recommend: AI is an excellent hypothesis generator and semantic grouper, but never a data source. Data must come from Search Console (for your domain), classic tools (for market volume validation) and Google directly (for real SERP and AI Overviews). Combine them well and you multiply productivity without losing rigor.
In our experience with SMBs in Barcelona, Sabadell and Girona, this hybrid approach produces lists that are 3 times more relevant and, above all, prevents the client from spending resources on keywords with no real commercial path. If you want help defining your keyword map with this method, we run a free diagnostic audit where we give you actionable priorities for the next 3 months.
7. Frequently asked questions
Can I do keyword research with only a free AI?
For an initial exploration yes, but not for final strategy. You'll miss out on volume and difficulty validation. I recommend combining AI with a classic tool (Ahrefs, Semrush or, at the very least, Keyword Planner with an Ads account).
Which AI is best for keyword research in 2026?
There is no clear winner. Claude and ChatGPT work better for clusters and intent; Gemini integrates Google results that help with AI Overviews; Perplexity is useful for seeing which sources Google cites. The combination beats any single tool.
How many keyword clusters does an SMB need?
In Catalan SMB projects, 10-15 well-developed clusters typically cover the customer lifecycle. More tends to dilute effort; less falls short if you want to rank across the whole funnel.
Can AI fully replace an SEO consultant?
No, not yet. AI accelerates tasks but strategic judgment, reading the local market and prioritizing by business needs are still human. The difference between a profitable project and a mediocre one is set by that strategic layer.
How do I know if a keyword is impacted by AI Overview?
The only reliable way is to run the search in an incognito window and see whether Google displays the AI summary. Tools like Semrush are starting to add this filter, but manual verification remains more reliable for searches in Catalan or Spanish in Catalonia.
Conclusion: AI doesn't replace keyword research, it amplifies it. If you're still using the workflow from two years ago, you're losing hours and, more importantly, opportunities. If you'd like us to review how to integrate AI into your current process without it becoming chaos, we run a free 30-minute diagnostic session and give you a concrete plan for your sector.