Direct answer: before an AI copywriter, social scheduler or influencer-briefing tool **publishes** an advert that references a flight-inclusive package (or its price) to UK consumers, the ATOL holder must already know whether ATOL Standard Term 1.3 applies — and, if it does, that the creative will contain a **proportionate reference to ATOL protection**. An LLM that “sounds compliant” does not create that gate.
That is the operator question Kaize is answering here. We are a UK AI product studio for travel and hospitality: we help tour operators, agents and hospitality groups put AI into marketing and booking stacks **with deterministic controls**, not hope.
Primary evidence is first-party UK Civil Aviation Authority material: CAP3236 Guidance in Support of ATOL Standard Term 1.3 (in force with AST 1.3 from **1 April 2026**) and the CAP3208 decision document that amended the term. Secondary reporting (for discovery only): Fox Williams’ summary of the 1 April 2026 change.
Why this is a marketing-stack problem, not a prompt problem
AST 1.3 states that ATOL holders that advertise ATOL protected services to UK consumers must, unless the CAA agrees an alternative approach that achieves equivalent consumer clarity, ensure that the advert contains reference to ATOL protection.
CAA guidance is clear that **not every holiday or flight advert is caught**. The decision turns on whether the advertisement makes reference to ATOL protected products — for example by showing the price of a flight-inclusive package or otherwise making specific reference to a flight-inclusive package / ATOL protected product.
- **In scope (typical):** social posts, digital display, short video, radio-style audio, and contracted influencer content that promote a flight-inclusive package or its price to UK consumers.
- **Often out of scope:** pure brand awareness, destination-led content with no package reference, hotel-only / accommodation-only standalone services.
- **Explicitly out of scope in the revised policy:** paid search advertising, including where AI is used to return search results.
- **Influencers:** only where there is a contractual relationship **and** the ATOL holder has editorial control — and only when they are advertising a specific ATOL protected product.
A generative model that drafts “7 nights Magaluf from £499 flights + hotel” for Instagram does not know your ATOL status, your product mix, or whether `#ATOLProtected` / logo / voiced line was injected. Those facts live in the **campaign brief, product catalogue flags and publish gate** — the same class of control Kaize argues for on AI agents that access booking systems.
How this differs from AST 1.4 booking-assistant disclosure
Kaize already covered what booking assistants must show **before a UK consumer chooses** an ATOL-protected option in ATOL AST 1.4 AI booking-assistant disclosure. That page is about the **purchase journey**.
AST 1.3 is about **advertising and promotional marketing** aimed at UK consumers. Different surface, different trigger, different artefacts (logos, tags, pinned comments, audio lines). Do not merge the two gates into one “ATOL prompt”.
What must already be true before AI generates the creative
1. Product classification is known
The item or offer the model is describing must be labelled in your catalogue or brief: ATOL-protected flight-inclusive package vs standalone hotel / flight-only / linked arrangement / brand-only. If the AI invents a package price from scraped fragments, you have already lost the classification gate.
2. Audience geography is UK (or treated as UK-aimed)
CAP3236 applies the requirement to advertising of an ATOL protected service **aimed at UK consumers**. Geo-targeted campaigns and UK-English social calendars should assume UK aim unless a deliberate non-UK audience control exists.
3. In-scope decision is recorded
Before publish, a deterministic check answers: does this creative reference a flight-inclusive package / ATOL protected product (including price)? If yes, AST 1.3 reference is required. If the creative is ambiguous such that consumers could infer a flight-inclusive package, CAA guidance notes you may wish to use wording such as “All flight-inclusive packages are ATOL protected,” and you may seek CAA advice — that is a **human compliance** call, not an unsupervised model guess.
4. Proportionate ATOL reference is injected for the format
- Text-only social: written words “ATOL protected”.
- Image + text: “ATOL protected” and/or ATOL protected logo; tags such as #ATOLProtected and/or pinned comments may also be used where space is constrained.
- Video with voice: written and/or logo reference, plus audible “ATOL protected” where a voiced statement exists (short clips may omit audio if only a few seconds — still keep written/logo).
- Audio-only: audible “ATOL protected”.
- Paid/commission influencer with editorial control: specify the reference in the brief and verify before go-live.
CAA’s case studies in CAP3236 are guidance, not an exhaustive checklist — but they are the right design input for your **format templates**, not something to “ask the model each time”.
5. Channel exclusions are respected
Do not invent an AST 1.3 duty for **paid search** creatives solely because an AI wrote the RSA headlines. CAP3208 / CAP3236 state the revised policy **does not apply** for paid search advertising, including where AI is used to return search results. Keep that exclusion explicit in the stack so operators do not over-block search while under-blocking social.
What changes when AI auto-publishes the ad
Unsupervised auto-publish is the failure mode. Typical breaks:
- Model invents a package price from a hotel-only rate file → suddenly in scope without ATOL reference.
- Influencer UGC pipeline has no contract/editorial-control flag → AST 1.3 influencer limb never evaluated.
- Short-form video template strips logos “for aesthetic” → proportionate reference removed.
- Green-claims or drip-pricing copy is generated in the same pass without the sibling controls on green claims and DMCC total price.
Treat ATOL reference injection like any other production control: fail closed when classification is missing; log the in-scope decision; keep a human path for ambiguous creatives.
Practical control pattern for travel marketing AI
- **Brief schema:** product_id, atol_protected (bool), product_class (flight_inclusive_package | standalone | brand | other), audience_uk (bool), channel (social|display|video|audio|influencer|paid_search), influencer_contractual_control (bool|n/a).
- **Pre-publish gate:** if audience_uk and references_atol_product and channel != paid_search → require atol_reference_artefact (text|logo|audio|tag|pinned_comment) matched to format.
- **Evidence pack:** store CAP3236/CAP3208 version links beside the campaign, not inside the prompt.
- **Sibling gates:** package-travel booking changes stay on the PTR booking-journey page; booking-time ATOL disclosure stays on AST 1.4.
What Kaize will and will not claim
We will not invent client results, rankings, or “AI that auto-complies with ATOL.” We will say: AST 1.3 is now a first-class **marketing publish control** for UK-aimed creatives that reference ATOL-protected flight-inclusive packages, and generative tools belong behind that control — the same philosophy as our production-break and operations automation pieces.
If you want a structured review of where AI belongs in marketing and booking journeys — and which compliance components must stay deterministic — start with Kaize’s AI Operations conversation, not a prompt library.
FAQ
Does AST 1.3 apply to every social post about holidays?
No. CAP3236 says not all advertising of holidays and flights is caught. Brand, destination or hotel-only creatives without reference to an ATOL protected / flight-inclusive package are generally out of scope. Specific package prices or clear package promotion to UK consumers typically bring the creative into scope.
Do AI-written Google Ads need “ATOL protected” under AST 1.3?
The revised AST 1.3 policy **does not apply** for paid search advertising, including where AI is used to return search results. Other advertising and consumer-law duties may still apply — do not treat the paid-search exclusion as a blank cheque for misleading ads.
What about influencers using AI captions?
AST 1.3’s influencer limb applies where there is a contractual relationship and editorial control, and the content advertises a specific ATOL protected product. AI-written captions do not remove the holder’s editorial-control duty — they increase the need for a checked brief.
Is this legal advice?
No. ATOL Standard Terms, CAA guidance and your licence conditions remain with your compliance and legal owners. This page explains what must be true in an AI marketing stack so operators do not confuse generative fluency with ATOL advertising compliance.