Continuously calibrated against X's open-source recommendation algorithm and our proprietary diagnostic systems. We surface what's actually moving distribution — right now, not last quarter.
xDoctor is a paired reputation-management product organized along the temporal axis of your relationship to your content.
Paste a draft. Get the predicted reach band, the negative-signal probability, the specific words and structural features that are dragging the score, and targeted edits. Each iteration becomes training data.
Every historical post the model flags as potentially harmful, surfaced for paced manual review. Visit on X, decide keep or delete, train the model with every decision.
Curated cleanup for: Reposts of harmful accounts. Posts that aged poorly. Replies to mass-reported conversations that brands might deem unsafe.
X open-sources its recommendation algorithm and ships updates roughly monthly. Most reach-optimization tools cite years-old thinking. We track every release the day it lands — see the widget in the top-right corner — and combine that public signal with proprietary diagnostic systems built specifically for X's classifier surfaces. The result is a model that's actually current.
xDoctor is currently in open Alpha testing. If we scale too quickly, we may have to close it to additional users for a time. Sign in below to be a part of it — read-only access, deletions through the X API only with your explicit approval, and we never post on your behalf.
sign in with X →No. The famous numbers — replies worth 13.5x a like, a report worth −369 — come from the 2023 open-source release, and the system that used them has been replaced. The current open-source X algorithm (released January 2026, updated May 2026) still combines predicted engagements with a weighted sum, but the 2023 values are obsolete and the current numeric weights are not published in the code. read the sourced answer →
The live system scores every post by summing nineteen named engagement predictions — the probability you will like it, reply, repost, expand a photo, watch a video to quality-view depth, share it three different ways, dwell on it, quote it, click through to the author's profile, follow the author — and four negative predictions: the probability you will click "not interested," block the author, mute the author, or report the post. The signal names are in the open at the pinned commit. The weight on each signal is not. read the sourced answer →
In the open-source X algorithm, a post reaches a viewer only by surviving three stages: it must be retrieved as a candidate (from followers via Thunder, or globally via Phoenix retrieval), survive ten named pre-scoring filters, and then score well enough in a ranking that explicitly subtracts for predicted blocks, mutes, and reports. Near-zero impressions means the post is dying at one of those three stages — and they have different causes. read the sourced answer →
Effectively, yes. The live system transcribes the audio of your videos with speech recognition and folds the transcript into the same representation it builds from your text. It embeds your images and your text together in one multimodal model, so a post's pictures are part of what the ranker understands, not decoration around it. And video duration is hydrated as a ranking feature with its own scoring gate. For creators who work in video and images, the algorithm is reading far more of your post than its caption. read the sourced answer →
The folk version — a secret switch that silently zeroes out an account — does not appear anywhere in the open-source X algorithm. What the code does show is several real, named mechanisms that reduce or remove visibility: a post-level visibility filter for spam and policy categories, per-viewer filters driven by blocks and mutes, safety classifiers, and negative ranking weights for posts the model predicts you would block, mute, or report. read the sourced answer →