Content
How do you turn the EU AI Act into a lead source for your software?
Copy for AI
Answer the EU AI Act questions your buyers already ask, one page per question.
That’s how the regulation turns into a lead source for your software.
Buyers find you at the moment they screen vendors.
The verdict: build an explainer, a “does our use case count as high risk?” page, a checklist, a deadline page and your own AI Act page for procurement teams.
Here’s the situation we see. Your product has AI features, and in tenders your sales team increasingly receives an AI Act questionnaire.
Marketing hasn’t got a single page on it.
This is how we use content marketing to turn a regulation like this into a source of pipeline.
How to keep this kind of content accurate and current is covered in regulation as a lead source.
What do software vendors need to know about the EU AI Act?
The AI Act sorts AI systems into four risk levels. Your buyer wants to know which level your feature falls into.
The four levels:
- Unacceptable risk: banned.
- High risk: strict requirements.
- Transparency risk (limited risk): mainly openness.
- Minimal risk: no new rules.
The timeline as of 28 September 2026
This is the situation as of 28 September 2026, according to the European Commission:
| What | From |
|---|---|
| AI Act enters into force | 1 August 2024 |
| Prohibited AI practices and AI literacy | 2 February 2025 |
| Rules for general-purpose AI models (GPAI) | 2 August 2025 |
| General application, with exceptions, and transparency rules | 2 August 2026 |
| High risk in certain sensitive areas | 2 December 2027 |
| High risk in products such as lifts or toys | 2 August 2028 |
The same dates on one line:
What the AI Omnibus changed
The high-risk dates were moved by the AI Omnibus.
A political agreement on it was reached on 7 May 2026.
The amendment entered into force on 27 July 2026.
You can follow every change on the Commission’s own page:
AI literacy changed too
According to the Commission, AI literacy remains an obligation for providers and deployers of AI.
But the Omnibus removed the requirement for a “sufficient” level.
Anyone deploying high-risk systems must still train the staff who carry out human oversight.
National market surveillance authorities supervise from 2 August 2026.
What do the transparency rules mean for your product?
They affect almost every software company with AI.
Three rules matter most:
- People who use a chatbot must know they’re talking to a machine.
- Certain AI content, such as deepfakes, must be visibly labelled.
- AI output must be marked in a machine-readable way.
For that machine-readable marking, systems already on the market before 2 August 2026 get a transition period until 2 December 2026.
Who has to do what
The questions and answers on Article 50 make clear who has to do what.
For a chatbot, the obligation lies with the provider, so often with you.
The notice comes at the start of the first interaction. The exception: an average, well-informed user can already see it’s AI.
Watch out: read that exception strictly.
Labelling a deepfake, on the other hand, is the job of the deployer of the system.
That’s often your customer.
Put the split in a table
Your buyer wants to know what you handle and what they do themselves. Put it in a table on your site:
| Obligation | Who | What you put on your site |
|---|---|---|
| People know they are talking to AI | Provider (you) | How and when your product tells them |
| Mark AI output in a machine-readable way | Provider (you) | Which marking you use and since when |
| Visibly label deepfakes | Deployer (your customer) | Which settings your customer gets for this |
Running a chatbot on your own site?
Our guide to an AI chatbot for a B2B website covers the setup side.
What questions do buyers ask software vendors about the AI Act?
Procurement, legal and IT ask the same seven questions:
- What AI is in your product, and which model does it run on?
- Which risk class does that feature fall into, and our use of it?
- What documentation do we get about the system, the data and the limits of the output?
- How do users see that they are working with AI?
- Can we switch off the AI feature or let a human decide?
- How do you help our team with AI literacy?
- Who follows up when the rules or the dates shift?
Put those answers on your site.
Then your buyer and their AI assistant will find them before they call.
That’s not a guess. AI assistants show exactly which pages they used:
Perplexity’s sources panel for our own test prompt, 17 October 2026: 10 web pages behind one answer.
Your AI Act page should be one of those sources. Our guide to GEO for SaaS shows how to get there.
Tip: questions 3 and 5 overlap with AI governance.
Our guide to AI agent governance and risks helps you phrase the answers.
Which pages do you build around the AI Act?
Pages where your product is the answer.
Don’t build a generic “what is the AI Act” page as your only piece.
Plenty of others already write that, including the Commission itself.
Six pages, three buying stages
| Page | Buying question it answers | Stage |
|---|---|---|
| The AI Act for [your sector], explained briefly | ”What does the AI Act mean for HR software?” | Awareness |
| Is this high risk? | ”Does AI screening of candidates count as high risk?” | Awareness |
| Checklist for deployers of AI | ”What do we need to do before 2 December 2027?” | Consideration |
| Deadline page with timeline | ”When do the high-risk rules apply?” | Consideration |
| Your AI Act page for procurement teams | The seven questions above, answered per feature | Decision |
| Comparison or selection page | ”What should you look for in software with AI features?” | Decision |
Sales mostly sends the procurement page.
For the comparison, follow our approach to comparison pages.
Add an FAQ block to every page
Short questions, short answers. They suit buyers and AI answers alike.
Mark them up as described in our guide to FAQPage schema. That’s how the FAQ at the bottom of this page is built too.
Further reading: the same playbook works for other EU rules.
See our page on European Accessibility Act compliance.
How much search volume is there around the AI Act, and what does a click cost?
More than you think. And the clicks are cheap.
These are average monthly Google searches over the last 12 months (Google Ads data, September 2026, cost per click in USD):
| Search term | Belgium | Netherlands | Cost per click BE / NL |
|---|---|---|---|
| ai act | 2,400 | 3,600 | USD 3.58 / 3.03 |
| eu ai act | 1,000 | 2,900 | USD 4.61 / 4.53 |
| ai verordening | 170 | 1,000 | not reported / USD 2.63 |
| ai geletterdheid | 90 | 1,000 | USD 3.62 / 4.12 |
Compare that with your category
A buying term costs far more: the median across 106 Benelux software terms is about USD 22 per click.
That’s roughly five to seven times the AI Act terms above.
Check yours in our CPC benchmarks by industry or work it out with the click-or-article calculator.
Long questions show no volume, but they’re searched
Specific questions such as “ai act risk classes” show no volume in the tools.
Yet people do search them. They just type them as long questions.
New pages get a bigger share of their impressions from queries of seven words or more than older pages do.
At Hedgehog that’s 19% versus 10%. At Kaizo, 16% versus 10%.
Our analysis of ChatGPT queries getting longer shows the same shift on the AI side.
Turn the questions into headings
So write those questions word for word as subheadings, as in answer-first writing for AI.
Question headings also earn the click more often on our own site:
Non-branded queries ranking in the top 10, customerimpact.be, 1 June to 15 October 2026. From our CTR study.
That’s 0.19% against 0.09%: about twice the click rate for question queries.
What is still changing in the AI Act?
The Omnibus shows the risk: published dates can go out of date.
Published a start date for high risk before the Omnibus? Then your page may now be out of date.
Fix: update the date on the same URL and add the real update date.
How to do that systematically is covered in content refresh. Why the visible date matters is in our guide to content freshness.
Three points to track
Track these on the sites of the Commission and your national supervisory authority over the coming months:
- 2 December 2026: the end of the transition for marking AI output.
- National implementation: how Belgium and the Netherlands organise supervision and penalties.
- The high-risk dates: currently 2 December 2027 and 2 August 2028, but check them with every announcement.
According to the Commission, December 2026 also brings a new ban on AI that creates non-consensual intimate images or child sexual abuse material.
Rule of thumb: each point is a news moment. Update your page within a week.
When are we not the right choice?
In three cases, and we’d rather say so up front:
- Your product needs a conformity assessment as a high-risk system. You need a lawyer first, not content.
- There’s no AI in your product and your customers don’t buy AI. This regulation will bring you little.
- You want guaranteed AI Overview placement. Nobody can honestly promise that, us included.
If it does fit
Content starts at 950 euros per month for 5 articles, with a minimum of three months.
The broader plan is in reducing your dependence on Google Ads. Our guide to SaaS lead generation shows where AI Act pages fit in your wider pipeline.
Want to know which AI Act questions your buyers ask most? Contact us and we’ll map them for your product.
Frequently asked questions
What is in the AI Act regulation?
The AI Act classifies AI systems by risk. Unacceptable risk is banned, high risk faces strict requirements and transparency risk mainly calls for openness. For minimal risk, the AI Act introduces no rules. On top of that come rules for general-purpose AI models and an AI literacy obligation.
Does AI screening of candidates count as high risk?
The Commission lists employment as one of the sensitive areas where high-risk rules apply, from 2 December 2027. Whether your feature falls under it exactly depends on what it does. Have a lawyer assess that, then write a page that explains that assessment to your buyers.
Do we need to label AI-generated text on our site?
Not always. The label applies to text you publish to inform the public on matters of public interest. Text that a human genuinely reviews or edits needs no label. A superficial or purely formal check does not count as editing, the Commission says.
Free website scan
Enter your website and get an automatic scan within minutes, with concrete technical and SEO improvements. No sales pitch.
We only use your details for your scan. No spam, unsubscribe anytime.