Wikimedia already publishes pageviews REST. The lock is not access. The lock is scale. A raw dump does not sit on the same 0-100 as google search or tiktok. Trends API takes source wikipedia and an article title. The live list is type Wikipedia Trending. The caller does not send a Wikimedia token. body is a JSON string and must be parsed a second time. Docs: the API reference. Wikipedia notes: the Wikipedia guide. Caps: pricing.
Wikimedia pageviews REST is public
That API is the right tool for a research notebook that wants project, access, and agent splits. It is the wrong tool when the warehouse grain is "one score per platform per week." Do not copy Wikimedia daily integers into a column named score and compare them to Google's 0-100.
The job is a 0-100 next to other sources
POST https://api.trendsapi.ai/api
Authorization: Bearer <api_key>
Content-Type: application/json
{"mode":"get_growth","source":"wikipedia","keyword":"ChatGPT","percent_growth":["3M"]}
A live pull on 2026-08-27 returned search_term Chatgpt, 3M growth +38.08 percent, recent_value 47.5, baseline_value 34.4, recent_volume 6574, baseline_volume 4760. volume_growth was +38.11 percent. Compare that row to a google search row only after both are on this product's scale.
source wikipedia, article title as keyword
Use the article heading, not a URL and not a page id. ChatGPT is a title. https://en.wikipedia.org/wiki/ChatGPT is not. If the inner payload is empty, the title is wrong or too obscure. Check the live article name.
Wikipedia Trending for the live list
{"mode":"get_top_trends","type":"Wikipedia Trending","limit":25}
No keyword. This is which articles are trending now. It is not the history for one title. get_time_series on source wikipedia is that history. Keep the board and the title series apart.