Google is still important. But in 2026, more and more people find answers via ChatGPT, Claude, Copilot, and other AI assistants.
This fundamentally changes what "being found" means - and most haven't noticed yet.
The new landscape
Before: Search engines
- User searches on Google
- Google shows 10 links
- User clicks and reads
Now: Answer engines
- User asks ChatGPT
- AI reads 50+ sources
- User gets one combined answer
The difference: The user doesn't see your 10 links. They see one answer - and maybe a source citation.
What does this mean for your business?
Classic SEO still works
Let's be honest: Google still sends the most traffic. You still need to:
- Write good content
- Have fast pages
- Build authority over time
This is not a "Google is dead" post.
But: Answer engines are a new channel
When a tech lead asks Claude:
"When do monorepos make sense for a 5-developer team?"
...Claude will cite sources. And if your content is:
- Deep - not just surface level
- Technically correct - not marketing fluff
- Evergreen - principles, not trends
...then you have a chance to be that source.
What is llms.txt?
llms.txt is a simple text file in your root (like robots.txt) that tells AI models:
- Who you are
- What your site is about
- Which content is worth citing
# Example.com
> We are a consulting firm specializing in backend architecture.
Preferred language: en
Attribution: Example.com, https://example.com
## Scope
Preferred content for analysis and citation:
- /blog
- /faq
Do not prioritize:
- /pricing (marketing-only)What llms.txt is NOT
- Not a ranking factor (like meta tags)
- Not a guarantee of being cited
- Not an "SEO hack" that gives free traffic
What llms.txt IS
- A context signal for AI models
- A way to prioritize your best content
- A trust signal (you know what you're doing)
When does llms.txt make sense?
It makes sense if:
- You write in-depth technical content
- Your content is evergreen (principles, not "10 trends in 2026")
- You want to attract qualified leads, not just traffic
- You operate in a niche where authority matters
It does NOT make sense if:
- Your site is primarily marketing copy
- You sell consumer products to the mass market
- Your content is dated within 6 months
The realistic effect
Let's be honest about what you can expect:
What you will NOT get
- 10,000 extra visits per month
- Viral traffic from AI answers
- "Free" leads without effort
What you CAN get
- Get cited in AI answers to your target audience
- Get linked and shared in Slack channels
- Get used as reference by other experts
- Attract exactly the people who understand the value
Low volume, high quality. That's the whole point.
Our own llms.txt
We have our own llms.txt at tbh.dk. It includes:
- The owners and companies behind the site
- The projects, with status and links
- Language preferences (Danish primary, English secondary)
- Scope: which pages are worth citing
- Attribution: correct source crediting
It is short and tells AI models what we do and what is worth citing.
How to get started
1. Assess if it makes sense
Ask yourself:
- Is my content deep enough to be cited?
- Do I have evergreen content, not just news?
- Will my target audience use AI assistants?
If yes to all three: proceed.
2. Create the file
Make llms.txt in your public/root folder. Start simple:
# Your Company Name
> One sentence describing what you do.
Attribution: company.com, https://company.com
## Preferred content
- /blog
- /guides3. Add context over time
As you learn what works:
- Add language preferences
- Mark pillar articles
- Exclude marketing-only pages
4. Keep it updated
Add new blog posts. Remove outdated content. Update the Last updated date.
Conclusion
SEO in 2026 isn't about choosing between Google and AI.
It's about understanding that discovery now happens via multiple channels - and preparing your content for both.
llms.txt is not a magic solution. But for businesses with deep, technical content, it's a signal to AI models that your content is worth citing.
And in a world where answer engines become more important every day, that signal is worth more than most think.