What's New in Content Recommendations
Discover the latest updates to Arc XP's Content Recommendations
Release dates
Sandbox and Production - Monday, July 13, 2026
Release summary
Signals API
Editorial Signals are a layer of human-controlled ranking rules that sit on top the Content Recommendation API’s machine-learned recommendations.
They let your newsroom override, nudge, or filter the model’s output without retraining anything: the changes take effect on the very next recommendation request.
In short: AI decides what is likely to engage a given reader; signals decide what is editorially encouraged, discouraged, or required for that reader to see (or not see).
Signals are exposed through a dedicated service: the Signals API (/signals/v1).
The four signal types
There are four distinct signal types. Each one answers a different editorial question.
Boost
”I want this to do better than the model thinks it will.”
Multiplies the item’s AI score by a weight between 1.0 and 10.0. Item rises naturally in the ranking.
Bury
”I want this to do worse than the model thinks it will.”
Multiplies the item’s AI score by a weight between 0.0 and 0.9. Item falls in the ranking but isn’t removed.
Pin
”This item must appear at position N.”
Forces the item into a specific 1-indexed slot, even if the content was not included in the preliminary candidate set. An absolute position override.
Exclude
”This item must not appear at all, regardless of relevance.”
Removes the item from the candidate set entirely. Takes precedence over every other signal, including Pin.
Boost and Bury are soft — they bias the ranking. Pin and Exclude are hard — they override it. The reranker applies them in a deterministic order (exclude → boost/bury → sort → pin → truncate), so the outcome is predictable no matter how many signals stack up on the same response.
Learn more in this Signals API guide.
Recommendations API: remove content anchor
The Recommendations API previously had the ability to anchor a recommendation based on the content ID. The initial implementation did not provide the expected results, and we are re-visiting the approach. For now, the suggested approach is to leverage anchoring by applying a section filter to the Recommendations API to achieve some measure of relatedness.
Recommendations API: Recency preview before ingestion
Before your catalog has been ingested and the model has trained, GET /recommend/v1/recommendations still returns a usable response so you can build and validate your rendering ahead of go-live.
In this state the API returns the five most recent active catalog items as a recency preview. The API returns up to five items in the free subscription tier. It is possible to request fewer items, but not more.
Preview items are fully enriched and carry the same attribution as personalized results, so your card-rendering code exercises the real response shape. What they do not carry is a relevance score — every preview item is scored 0.0. A response whose items are all scored 0.0 is the signal clients use to distinguish a preview from personalized results: no reranker is run, and no editorial signals (boosts, buries, pins) are applied to a preview.
If the catalog has no content yet (nothing has been ingested at all) there is nothing to preview, and the request falls through to the existing 403 response.
CDP Integration Guide
The Developer Portal now includes a new Connect Your CDP to the Recommender API guide that covers how to connect your customer data platform, analytics tool, or homegrown source into the Content Recommendations API.
This article outlines ingestion requirements and best practices along with common integration patterns covering:
BlueConic
HTTP-based connectors like Amplitude, Adobe Real-Time CDP, and Tealium
S3-transport connectors like Permutive and ActionIQ
Analytics-only stacks like GA4 and Sophi
Users affected
Beta customers
Action required
None
Release dates
Sandbox and Production - Tuesday, May 26, 2026
Release summary
Stronger recommendations for anonymous readers
The Content Recommendations API now uses anonymous reader activity more effectively to improve recommendation quality.
Improved multi-site onboarding
Multisite customers can now roll out the Content Recommendations API site by site instead of importing every site at once.
Improved API validation and feedback
The API has been updated to perform more stringent validation against the schema and provide clear feedback for rejected requests.
Users affected
Beta customers
Action required
None
Release dates
Sandbox and Production - Monday, May 18, 2026
Release summary
Editorial Signals are a layer of human-controlled ranking rules that sit on top of the Content Recommendations API's machine-learned recommendations.
They let your newsroom override, nudge, or filter the model's output without retraining anything — the changes take effect on the very next recommendation request.
In short: AI decides what is likely to engage a given reader; signals decide what is editorially required for that reader to see (or not see).
Signals are exposed through a dedicated service — the Signals API (/signals/v1).
The four signal types
There are four distinct signal types. Each one answers a different editorial question.
Signal type | Editorial question | Effect on the ranked list |
|---|---|---|
Boost | "I want this to do better than the model thinks it will." | Multiplies the item's AI score by a weight between 1.0 and 10.0. Item rises naturally in the ranking. |
Bury | "I want this to do worse than the model thinks it will." | Multiplies the item's AI score by a weight between 0.0 and 0.9. Item falls in the ranking but isn't removed. |
Pin | "This item must appear at position N." | Forces the item into a specific 1-indexed slot, even if the content was not included in the preliminary candidate set. An absolute position override. |
Exclude | "This item must not appear at all, regardless of relevance." | Removes the item from the candidate set entirely. Takes precedence over every other signal, including Pin. |
These controls are currently API only. A user experience will be added in an upcoming release.
Users affected
Beta customers
Action required
None
Release dates
Sandbox and Production - Monday, May 11, 2026
Release summary
Editorial Signals are a layer of human-controlled ranking rules that sit on top of the Content Recommendations API's machine-learned recommendations.
Content teams can now pass
sectionandcontent_typeparameters directly in theGET /recommendations/v1/recommendationscall.Items that don't match are filtered out before scoring begins — they never compete for slots and never appear in the response.
For example, you can now optimize the Sports module on your homepage to re-rank the items based on an individual consumer's interest.
Users affected
Beta
Action required
None
Release dates
Sandbox: May 04, 2026
Production: May 04, 2026
Summary
We are excited to announce the Open Beta of the Content Recommendations API, Arc XP's new personalization platform for media companies. Participating customers will receive six months of free access to the API (and an IFX integration used to ingest content) and can use it to power personalized site, app, and email experiences that drive deeper engagement with their audience.
This doc summarizes what's available today, what it takes to get started, and what's on the roadmap.
What's available in the initial release
The Content Recommendations API is a personalization service that learns from your content catalog and audience behavior to return ranked content recommendations on demand. Each customer operates with their own fully isolated recommendation model: content, audience data, and the trained model are never shared across tenants.
The initial release supports two core recommendation types:
Personalized recommendations for a user overall — given a user identifier, the API returns a ranked list of content items tailored to that user's interaction history. Ideal for a “For You” module on a home page or a personalized section in a newsletter.
“More like this” recommendations in the context of a specific piece of content — given both a user and an anchor article, the API returns items similar to the anchor but still personalized to the reader. Ideal for an end-of-article or in-article recirculation module.
Cold-start behavior is handled automatically. New or anonymous users receive popularity-weighted results until they accumulate interaction history, and newly published content is eligible for recommendations immediately based on its metadata, no special handling required on the customer's side.
What participating customers need to do
The API is the personalization backend only. To put recommendations in front of readers, each participating customer is responsible for two integration workstreams:
Connecting their Customer Data Platform (CDP) to send behavioral events (page views, clicks, shares, engaged reads, scroll depth, and similar signals) to the Collector API. A richer event stream produces stronger personalization. More information about the Collector API can be found here.
Building the front-end UI that surfaces recommendations to readers — for example, a “For You” module on the home page, a recirculation rail on article pages, or a personalized section in an email template. The API returns ranked content IDs; the customer hydrates them from their CMS and renders the experience. More information about the recommendations API can be found here.
Arc XP CMS customers can use an IFX recipe to wire up the content webhook end-to-end without custom code. Non-Arc XP CMS customers will need to build a direct integration from their CMS to the Collector API's content endpoint.
A full onboarding checklist is available on the Arc XP for Developer site.
What's coming next
This initial release is only the foundation and we have significant functionality slated for release over the coming weeks and months. Planned updates include additional filtering and personalization controls, stronger editorial controls, and deeper integration hooks. Customers who join the beta now will benefit from those additions as they ship, within their six-month free access window.
Getting Started
Talk with your Technical Account Manager about enrolling in this beta. The free access beta is limited to customers who begin their beta engagement in May 2026.