Back to blog

X Algorithm Update: How Multi-Brand Accounts Can Test Content Around Reading Time

In an October 2026 announcement, X described a model that better predicts time spent and retrieval trained to surface posts people read through. This guide shows brand teams how to turn that direction into content tests, measurement, and organized account workspaces with PurpleMark.

The same topic may receive scattered engagement on one brand account and spark an extended discussion on another. Teams running several brands can blame this on another algorithm change, then alter posting time, length, and format all at once—making it hard to tell what helped.

On October 20, 2026, Elon Musk described two directions in an announcement about timeline recommendations: a new ranking model intended to predict likely time spent more accurately, and a retrieval model trained to find posts people read all the way through. When checked, the post showed about 3.173 million views. This is a snapshot of one post, not a measure of search volume for the topic.

For operators, the announcement suggests a question worth testing: do our posts give the intended readers a reason to keep reading?

Turn the announcement into a testable question

Retrieval finds content that may enter the recommendation pool; ranking orders the candidates for display. The public announcement did not provide a traffic formula or the specific weight of each signal.

The ideas below are operational suggestions based on the announced direction and should be checked against your own account data. Length, cadence, and interaction format should fit the audience. A short post that explains its point fully can still make for a clear reading experience.

Choose a question a brand has had to answer repeatedly—for example, who a product is for, how to complete an operation, or why a common misconception arises. Make it specific so you can judge whether the post reached the right people.

Make the opening deliver on the reader’s expectation

A post’s first sentence can identify the audience, the problem, and what readers will get by continuing.

“We collected three marketing tips” is broad. “Managing X accounts for three brands? How can you make sure a reply goes to the right customer? Here’s an example handoff log” helps readers decide whether the post is relevant.

Continue from the question in the opening. Give the needed takeaway first, then support it with an example or steps; explain necessary terms where they appear. Close by inviting readers to share a relevant situation so the discussion stays on topic.

The aim is clearer information. Run a small experiment and see whether it brings more relevant replies, account visits, or business inquiries.

Design one content experiment for one brand

Choose a topic, record the question for this round, and change only one main factor. This is a suggested experiment design, not a ranking rule published by the platform.

QuestionApproaches to compareWhat to review
Do readers understand the subject?A broad opening vs. one tied to a clear use caseDo replies address the intended question?
Is the explanation easy to follow?One long block vs. paragraphs ordered by informationDo replies ask again about points already explained?
Does an example help?State an opinion vs. add one concrete caseDoes discussion include more specific use cases?
Is the next step clear?A general invitation vs. one related questionAre there relevant inquiries or useful discussion?

Compare similar topics and observation windows where possible, and note confounders such as breaking news or paid campaigns. Audiences differ by brand, so establish a separate baseline for each; treating Brand A’s performance as Brand B’s expectation can cloud judgment.

Separate visible metrics from the conclusions you want to test

“People saw the post,” “people read it through,” and “people want to learn about the brand” are different questions. Keep the platform’s actual metric names and measurement scope in your notes.

Start with post views and the substance of relevant replies to assess whether the intended audience was reached. If the account dashboard provides profile visits or link clicks, include those too. Views alone do not equal a completion rate; do not infer a metric you cannot observe from a single number.

A short log supports ongoing reviews: brand and account, post link, publication time, hypothesis, main change, observation window, actual metrics, and the next adjustment. Review these notes before deciding whether to keep a particular approach.

Organize workspaces for a multi-brand team

When one operator handles several brands, PurpleMark professional antidetect browser lets the team group browser profiles by brand, create a separate profile for each business account, bind the account, and set frequently used startup pages. Fingerprint settings, cookies, and local storage are managed separately for each profile, helping maintain the corresponding sessions. Include the brand, platform, and region in profile names to make them easier to identify before opening.

For collaboration, arrange responsibilities using members, roles, permission groups, profile sharing, and operation logs. The person publishing enters the relevant brand profile; the person reviewing performance organizes notes by post link. During handoff, share the content hypothesis, published link, and next step. This keeps account workspaces clearly connected to experiment records.

How to start the next post

Choose one brand, one real question, and one main change. After writing, check that the opening states the value accurately, the body answers the question, and the ending offers a relevant next step.

Teams already using PurpleMark can open the client and organize the profiles and member permissions for this round. New teams can download the PurpleMark client, begin with one brand account, and expand the workspace as needed.

Information checked on November 11, 2026. The announcement documents a public change; the content experiment and team organization ideas are practical suggestions. Validate results over time against actual account feedback.