Source diversity in a single logged-out YouTube search — an exploratory pilot (n = 13)
This pilot asks a narrow, descriptive question: in a single logged-out YouTube search for a politically sensitive query, how do the top-ranked results split between legacy/mainstream outlets and independent/digital-native channels? It is a probe, not an audit — one query, one moment in time, thirteen results — so it describes a snapshot and cannot support causal claims about how “the algorithm” behaves in general.
I ran a clean, unauthenticated (logged-out) search for bangladesh politics on YouTube. Because the session was logged out, no personalization or watch history is in play — so this captures YouTube’s default search ranking for an anonymous user, not personalized recommendations. I read the ranked results from the ytInitialData payload (rank order, channel names, view counts) and coded each channel into one of two categories: Legacy/Mainstream Media or Independent/Digital Native. Coding was done by a single coder against an informal rubric; there is no second coder or formal inter-coder reliability yet, and some cases are genuinely contestable.
Of the 13 top-ranked results, 7 (53.8%) came from legacy/mainstream outlets and 6 (46.2%) from independent/digital-native channels. This describes one snapshot; it is not a measured tendency of the platform.
In this snapshot, legacy/mainstream outlets held 7 of the 13 top slots (53.8%), even though some independent channels posted large individual view counts (e.g., a single Dhaka Xpress video with over 81,000 views). That pattern is weakly consistent with a story in which ranking tracks institutional domain authority more than raw engagement — but the margin is a single video. With n = 13, reclassifying or re-ranking just one result flips the majority, so this is a hypothesis to test at scale, not a finding. It does not establish that YouTube “acts as a gatekeeper.”
| Rank | Channel | Video Title |
|---|---|---|
| 1 | ATN News Live | সরকারের ব্যর্থতায় রাজপথে কর্মসূচি: জামায়াত আমীর | Jamaat Amir | Bangladesh Politics |
| 2 | Daily Jugantor | আ.লীগের নাশকতার ছক, ৫ স্থানে চোখ প্রশাসনের | Awami League | Bangladesh Politics |
| 3 | Ekattor TV | বিরোধী দল কি সহনশীলতার বিপক্ষে অবস্থান নিচ্ছে? | Bangladesh Politics | Opposition |
| 4 | Daily Jugantor | ইপিজেড চাইলেন এমপি, উত্তরে যা বললেন স্বরাষ্ট্রমন্ত্রী | Salauddin Ahmed | Bangladesh Politics |
| 5 | Politics Tv | ভারতের সাথে বাংলাদেশের বাণিজ্য ঘাটতি বেড়েছে, মন্ত্রীর স্বীকারোক্তি | Politics Tv |
This is a single-query, single-moment probe with n = 13, coded by one person. It cannot speak to personalization (the session was logged out), to other queries or languages, or to change over time. YouTube search results are non-stationary, so these 13 results are a frozen record rather than a reproducible output. A defensible version would need multiple queries sampled repeatedly over time, a written codebook, and a second coder. Read the numbers as an illustration of the method, not as evidence about the platform.
The raw CSV, the scraping method, and the analysis code are public. One caveat on “replication”: because YouTube search results change constantly, rerunning the code returns today’s results, not the 13 captured here. The CSV is the archived record of this snapshot; the code reproduces the method, not the numbers.
View Code & Dataset on GitHub →