AI Assistant
Turn ads into conversations.
AI Assistant is Friendly Tags technology for AI-powered conversational advertising. It turns a standard placement into an interactive experience where audiences can ask questions, explore products or content, and move toward a relevant action without leaving the creative. This page covers what the capability can do and what a campaign team needs to provide to launch one. Available capabilities may vary by campaign, by the selected AI provider and model, and by configuration.
Animated example of an AI Assistant inside an ad. A shopper asks whether the advertised earbuds are good for running, the assistant answers and shows a product card, and the shopper asks a follow-up about colours. The brand, product and answers are fictional.
Brand, product and responses are fictional.
An ad that can answer questions.
What the capability is, who it is for, and how a conversation moves from a question to an action.
Instead of delivering a single fixed message, the creative can respond. An audience member types a question or selects a suggested prompt, and the experience answers using brand and product information the advertiser has approved.
Friendly Tags provides the underlying technology and experience layer: the creative behaviour, the connection to a supported AI provider, the grounding in approved content, the controls, and the reporting. Campaigns may be run directly by an advertiser or brand, or through an agency, publisher, creative partner or technology partner.
Built for results
- Natural, real-time conversation inside the ad unit
- Product, service and brand discovery through approved information
- Contextual recommendations and rich responses where configured
- Clear paths to relevant calls to action and landing destinations
- Actionable insight into what audiences are asking about
How it works, at a glance
Ask
The user types a question or selects a suggested starter prompt.
Understand
The connected AI model interprets the intent and identifies the relevant configured information.
Answer
A response is returned as text or, where configured, as a richer card with images, details and links.
Act
The user follows a call to action, explores further, or continues the conversation within the configured limits.
Learn
Available interaction data shows what audiences asked and where they engaged.
Your AI, your account
Where applicable, the experience connects to a supported AI provider or model through your own provider account and relationship. Provider usage, billing, data handling and service terms remain subject to your agreement with that provider. Setup covers how the connection is configured.
What we need to build your AI Assistant.
Nine inputs. The more complete and current they are, the better the experience performs.
Approved knowledge and content
FAQs, product and service information, approved documents, website content, offers, campaign messaging, or other material the experience is allowed to use.
Product or content feed
Structured data used for dynamic recommendations or rich response cards. Only needed if the campaign uses them.
Brand guidelines
Voice, tone, terminology, required messaging, prohibited claims, and any other brand rules the experience should follow.
CTA and landing URLs
The destinations users should reach when they are ready to learn more or take action.
Suggested starter questions
Ideally three or four high-value prompts that show users what they can ask.
Conversation settings
Preferred limits, timeout behaviour, fallback behaviour and usage considerations.
AI provider requirements
Your preferred supported AI provider or model, and the account or credentials needed for the connection. Subject to that provider's service terms, billing, policies and technical requirements.
Privacy and disclosure requirements
Any approved legal, privacy, consent or policy language that needs to appear.
Success metrics
The KPIs that define success for the campaign.
How the AI connection works
The experience connects to a supported AI provider or model through the campaign configuration. Where applicable, you supply and own the provider account and relationship, and Friendly Tags configures the conversational experience around your approved content, conversation behaviour, creative presentation, usage controls and campaign requirements.
Your AI. Your account. Your terms.
Because the experience uses your provider connection, provider usage, billing, data handling and applicable service terms remain subject to your agreement with that provider. Friendly Tags does not provide or replace the underlying third-party AI service.
Providers are not interchangeable
Not every AI provider or model supports identical functionality. Availability, capabilities, limits, latency, integration requirements, pricing and service terms can differ. A preferred provider or model should be technically validated before a campaign is scoped.
Knowledge, grounding and feeds
The experience is grounded in the information you approve. Response quality depends heavily on how clear, accurate and current that information is.
- Use current, approved brand information
- Remove outdated or contradictory material
- Give clear product and service names, descriptions, differentiators, offers, eligibility rules and restrictions
- Use a structured feed where dynamic product or content recommendations are needed
- Define the subjects or claims the experience should not address
- Update the source content whenever campaign information changes
Better input, better answers, better experience.
What the creative can do, and how you control it.
Capabilities and controls may vary by campaign, by provider and model, and by configuration.
Experience and creative capabilities
Free-text questions
Users can type their own questions.
Suggested questions
Starter prompts guide users and make it easier to begin.
Multi-turn conversation
Users can ask follow-up questions within the configured session and conversation limits.
Rich response cards
Responses may include text, images, product or content information, descriptions and CTA links where supported.
CTA links
Responses can guide users toward relevant destinations and campaign actions.
Idle and re-engagement messaging
A follow-up prompt can appear after a configured period of inactivity.
Fallback states
A defined message or next action can appear when a limit is reached or an appropriate answer cannot be returned.
Product and content feeds
Where a campaign needs the experience to surface specific products or content, a structured feed supplies the items. A feed can support dynamic recommendations, richer response cards with images and details, and reporting on which items were surfaced or clicked, where supported. A feed is optional: an experience can run on approved knowledge content alone.
AI usage and conversation controls
Conversations consume resources on the connected provider account. Controls balance the user experience, the campaign goals and provider usage. Available controls may vary by provider and configuration.
- Conversation or session limit
- Controls how much interaction is available to a user or session.
- Message limit
- Caps the number of messages or turns in a conversation.
- Token or usage limit
- Where supported and configured, helps manage AI consumption and campaign cost.
- Idle timeout
- Defines what happens after a period of inactivity.
- Fallback behaviour
- Defines the experience when a limit is reached, when the configured content cannot answer, or when another configured condition occurs.
What you learn, and how it stays on brand.
Reporting turns real conversations into direction for the next one. Safety starts with the content you approve.
Analytics and conversational insights
The experience can connect to reporting that shows how audiences engage and what they ask. The exact reporting set should be confirmed for each implementation.
Sessions and users
Conversation sessions, unique users where available, and time engaged.
Messages and turns
Messages sent and average conversation depth.
Provider and model usage
Configured provider or model usage metrics, including token-related usage where available.
Feed interactions
Relevant content or feed matches, surfaced items and link activity where supported.
Top questions and topics
Common questions, themes, keywords and audience interests.
Performance
Response activity, CTA clicks, engagement and other agreed campaign KPIs.
Insights can be used to refine starter prompts, knowledge content, CTAs and campaign strategy over time.
Privacy, security and brand safety
User disclosure
Users should be informed that they are interacting with an AI-powered experience.
Conversation and data handling
Where interaction data is recorded or retained, appropriate disclosure and approved policy language should be used.
Brand safety
The experience should be grounded in approved content and configured brand guidelines.
Data privacy
Handling and retention should align with the applicable campaign and client privacy requirements.
Security
Provider connections and data handling should follow the technical security requirements of the implementation.
Retention controls
Where applicable, define how long conversation and analytics data should be retained.
Requirements vary, so confirm them before launch
Privacy, consent, security and retention requirements can differ by client, provider, campaign and jurisdiction, and final requirements should be approved before launch. This page is a product overview and is not legal advice. See the Privacy Policy and Terms of Service.
From idea to launch.
What to expect, what works well in practice, and what to know before the campaign goes live.
The launch process
Discovery and goals
Define the campaign objective, audience, desired action, AI use case and success metrics.
Provide content
Collect approved knowledge, feeds, brand rules, CTA destinations, privacy requirements and starter questions.
Configure the experience
Set the provider or model connection, grounding, conversation rules, limits, fallback behaviour and campaign behaviour.
Build the creative
Design and develop the prompts, response presentation, cards, CTAs and the relevant states.
QA and response testing
Test creative behaviour, grounding, expected and unexpected questions, limits, fallbacks, links and provider behaviour.
Review and approval
Approve content, behaviour, disclosures and creative.
Launch
Deploy the approved campaign experience.
Measure and optimise
Review the available insights and refine content, prompts, CTAs or configuration where appropriate.
Best practices
- Start with a focused use case instead of trying to make the experience answer everything
- Provide clear, accurate, approved and up-to-date content
- Use three or four high-value starter questions
- Keep answers concise and easy to scan in an ad environment
- Use CTAs strategically, so they guide the next step without interrupting a useful conversation too early
- Set sensible conversation and usage limits based on the campaign objectives
- Define fallback messaging before launch
- Test common, unexpected, sensitive and out-of-scope questions
- Review audience questions and engagement insights after launch, and optimise
- Keep knowledge sources and feeds current throughout the campaign
Limitations and things to know
- AI responses are generated dynamically and are not identical to predetermined ad copy
- Response time can vary depending on the question, the provider or model, connectivity and configuration
- Answer quality and relevance depend on the content and instructions used to configure the experience
- Conversation, message and AI usage limits may apply
- Third-party AI provider availability and service behaviour can affect the experience
- Not every provider or model offers the same capabilities or integration options
- Rich cards, feeds, recommendations and advanced behaviour depend on the campaign setup and the available data
- The experience should be thoroughly tested before launch, including expected questions, edge cases, fallback behaviour, CTAs and brand-sensitive topics
- An AI Assistant experience should not be positioned as a replacement for human customer support unless that use case is specifically designed and approved
Frequently asked questions.
Can we use our preferred AI provider?
Where that provider or model is supported for the campaign, yes. The connection uses your own AI account where applicable, so usage remains subject to that provider's billing, policies and service terms.
Who owns the AI account and provider relationship?
You do. Friendly Tags connects the conversational experience to your supported AI provider and account; the provider relationship, billing, policies and applicable service terms remain with you.
Can users type their own questions?
Yes. The experience supports free-text questions as well as suggested starter prompts.
Can the experience use our brand or product information?
Yes. Approved brand and product content is a core part of configuring and grounding the experience.
Can it recommend products or content?
It can be designed to surface relevant products or content where the required feed, content and campaign configuration are available.
Can responses include images and links?
Rich response formats can include images, details and CTA links where supported by the implementation.
Can we control how long users can chat?
Yes. Conversation and message limits, and other usage controls, can be configured where supported.
How do we control AI costs?
Conversation, message and AI usage controls can help manage consumption. Exact cost management depends on the provider or model and the campaign setup.
What happens if a user stops interacting?
The experience can use a configured idle or re-engagement state.
What happens when a limit is reached?
A configured fallback or handoff message can guide the user toward the next action.
Can we see what people are asking?
Reporting can support questions and topics alongside other interaction metrics. Exact availability and retention should be confirmed for each campaign.
Is every answer pre-approved?
No. Responses are generated dynamically. Brand safety relies on approved source content, instructions, guardrails, testing, and provider or model configuration.
See it on your campaign.
Tell us the objective and the content you have, and we will walk through what an AI Assistant experience would look like for it.