Ads inside AI conversations are a format under testing: some assistants experiment with or announce advertising associated with answers. Details change fast and vary by service and market, but the essential for the advertiser is to prepare with clear content, identification rules and own measurement.
The essentials in a definition
Ads inside AI conversations are advertising formats integrated into assistants’ answers or accompanying them. They are under testing, change fast and require clear identification (UCPD), transparency in AI use (AI Act) and quality content, per guidance on AI-generated content.
What is known and not known
AI assistant providers and search platforms have tested or announced advertising formats integrated into answers or alongside them, with rules, availability and measurement that change. So avoid planning on rumours and always confirm current conditions with providers.
The stable principle is that advertising must be identifiable as such and not mislead the user.
Identification and trust
The Unfair Commercial Practices Directive requires commercial content to be clearly identified. In a conversational context, where the answer seems like a neutral recommendation, distinguishing answer from ad is even more important. The AI Act adds transparency obligations for certain uses of AI systems.
Avoid formats that mix ad and answer in a confusing way.
Preparing for a channel still being defined
- Follow official ads and conditions
- Prepare clear content and data
- Define identification and safety rules
- Test with a hypothesis and limited budget
How to prepare
Prepare with what you already control: clear pages that answer real questions, consistent product data, coherent brand messages and guidance on AI-generated content. Define which questions and decision moments matter to the brand, for when targeting options appear.
Do not promise results or commit large budgets before seeing data.
In a new format, transparency and your own data are what you can control.
Measure and test
If you get access to a test, define hypothesis, budget, deadline and indicators (qualified clicks, conversions, lead quality) and compare with known channels. Record measurement limits and learnings. Protect brand reputation with rules on contexts and sensitive topics.
Review the offer every quarter, because the landscape changes fast.
Decision table
| Aspect | What to prepare | Caution |
|---|---|---|
| Content | Clear, factual pages | Generic texts |
| Data | Consistent product and price | Divergent information |
| Identification | Ad clearly distinguished | Mixing with the answer |
| Measurement | Own hypothesis and indicators | Relying only on provider reports |
Common mistakes to avoid
- Planning on rumours. Conditions change.
- Betting a large budget early. Data are still limited.
- Ignoring identification. It may be misleading.
How to measure the result
When there is a test, track qualified clicks, conversion, lead quality and cost, comparing with known channels.
Limits and exceptions
This is a very recent, changing landscape, with limited official information and differences between providers and countries. Some features may not reach your market. Avoid commitments and always confirm current conditions.
Illustrative example
Illustrative example: a software company defines the ten decision questions customers ask, improves the pages that answer them, aligns product data and prepares a test hypothesis. When a provider opens limited access, it tests with a small budget and compares lead quality with other channels.
In practice: checkpoints
Follow official announcements and keep a short list of your questions for providers. Review brand-safety rules, prepare creatives that can work in text-led formats and plan a test budget in advance. Remember that conversations are personal, so respect context and avoid anything that feels intrusive.
Frequently asked questions
Can I already advertise in AI answers?
It depends on the service and market; confirm current conditions with the provider.
Should I wait?
Prepare content and data, and test only when there is access and clear conditions.
How do I avoid being misleading?
By identifying the ad clearly and keeping claims true.
Which formats should I expect?
They vary; follow official documentation instead of assuming.
Does this replace paid search?
For now it is a channel under testing; it complements and does not replace what works.
Prepare clear content and data, follow official conditions, require transparent identification and test with a hypothesis and limited budget.
Sources and recommended reading
- Google Search Central: generative AI content guidance
- Unfair Commercial Practices Directive (2005/29/EC)
- Artificial Intelligence Act (AI Act)
Last editorial review: October 2026. Informational content; always confirm current conditions in the official sources before deciding.