Autoposting
News goes to your channels on its own, rewritten or reposted.
Autoposting turns the stories in your project into posts in your channels, with no editor at the screen. It is a content factory, or auto-posting: a Telegram or MAX channel that fills itself from your sources, in your style, right away or on a schedule. An AI model writes one news item from all sources of a story, or the original post is forwarded as it is.
Open Autoposting («Автопубликация») at /autopublish. There are two modes: Online («Онлайн») posts each story separately as soon as it forms, and Scheduled («По расписанию») posts at set times, either as a topic digest or as the top story.
Responsibility. When you turn on autoposting, even text only, with no images and no AI rewriting, you instruct the service to post in your channels on your behalf without checking each post. Such posts count as published by you, and you are responsible for their content and consequences. AI texts and images may contain mistakes. TeleMax is a technical tool and is not responsible for what is published. To see posts before they go out, turn on review before publishing. Details are in the terms of service (sections 6.4–6.7 and 7.2) and the legal information.
Online mode publishes to Telegram and MAX. A VK channel currently receives only digests and posts published manually from the feed.
#How it works
- Story. Posts from sources about one event are grouped into a story.
- Source threshold. The story waits until the required number of independent sources have written about it. The number is set in project settings. Channels of one newsroom count as one: a post the outlet put on both Telegram and VK will not pass a threshold of 2.
- Delay. While sources keep writing about the story, publication is put off, so that the post is complete and not based on the first message. We never wait longer than the maximum. See When a post goes out.
- Selection. If the monitoring keyword filter is on, the story must contain a selected keyword. If the AI editor is on, it decides whether the story fits the channel's topic. A story that does not fit gets "Skipped" with a reason, and no text is written for it. The check itself uses the monthly token limit: about 200–300 tokens per story.
- Advertising. Stories marked as advertising are not published if "Filter advertising" is on.
- Token limit. The monthly token limit is shared by rewriting and the AI editor. If it runs out, the story is skipped: raw source text is never sent to the channel. You get a warning at 75, 90 and 100 % usage.
- Rewrite. The AI model writes the post from the facts of all sources, in the chosen post style.
- Safety check. The service checks the text: an empty answer, service words like "SKIP", a refusal or an explanation instead of a post never go to the channel.
- AI editor review of the post. If the editor is on, it rereads the finished post: is it a refusal, is it about this story, does it contain facts that are not in the sources. If the post is over the limit, the editor shortens it.
- Review before publishing. If "Show me the post before publishing" is on, the finished post first comes to you, in Telegram, MAX or on the site only, and waits for your decision. See Review before publishing.
- Sending. The post goes to all enabled channels with an image from the source, a list of sources and a signature. If "Post image" is set to generate an image, the AI model draws it after the post is checked. See Story images.
In Repost mode, right after selection, the first post from a source is forwarded as it is. Steps 5–9 are skipped; review before publishing (step 10) still applies. Stories marked as advertising are filtered out already at the source threshold.
Every decision (published, skipped or error) goes to the publishing log with the step, the reason and the AI model's text. The story status ("Publishing in N min", "Published", "Skipped") is also shown on its card in the feed. You can cancel publication there too.
#Review before publishing
The "Show me the post before publishing" («Показывать мне пост перед публикацией») switch decides who clicks "Publish": you or the service. On: the post is prepared as usual but goes to the channel only after your decision. Off: posts go out on their own.
The switch is on both tabs, and each tab has its own: on Online it is in the online mode settings, on Scheduled it is under the release schedule. For example, you can review digests and let the story stream go out on its own. A change takes effect after Save («Сохранить»).
Review is on for new settings. In projects where autoposting was set up before November 2026 it is off, so that posts do not start waiting without warning. Turn it on yourself if you need it.
When a post comes for review. At the same time it would have gone out on its own: delay, extension and maximum wait work as usual, and for scheduled releases it is the release time. Selection, the AI editor and the text check happen before you see it: you get exactly what would have gone to the channel. In Repost mode you get the post that would be forwarded.
Where to send it. To Telegram or MAX, in the same chat with the TeleMax bot that receives Monitoring alerts, or "Site only". If the bot is not connected yet, there is a "connect" link next to it. The "Send an example" button sends a sample message so you can see how it looks.
What you get. One message: the post text as it will appear in the channel (with the image, if the post has one), the number of attachments, the channels it will go to, and the time until which we wait for a decision. Below it are four buttons:
- Publish («Опубликовать»): the post goes to the channels within a minute, and the message changes to "Published".
- Queue («В очередь»): the post goes out later, at the next free slot in the queue (see below). The message changes to "Queued: goes out tomorrow at 09:00".
- Don't publish («Не публиковать»): the post does not go out, and the story gets "Skipped".
- Edit on the site («Изменить на сайте»): opens My feed with the story window. The post text is already in the editor, attachments and channels are ticked. Edit it and click "Publish". For a digest, the button leads to the Scheduled tab, which shows the assembled digest with its cover and the same buttons. If the post is from another of your projects, the site switches the project for you.
On the site. Posts waiting for a decision are shown in My feed in the Publications («Публикации») block above the stories, with the same buttons. Their number is shown next to the My feed menu item. The same block has an Upcoming («Скоро выйдут») list: stories queued in online mode without review, with their time and a "Don't publish" button. Turn on the browser notification "Posts to review before publishing" in settings, and the browser will also tell you about a new post.
How long we wait. The deadline is set in the online mode settings: 3 hours (default), 6 or 12 hours, a day, 3 days or no deadline. If there is no decision in time, the post is not published ("Time ran out"). A scheduled digest or top story waits until the next release, but no longer than 12 hours.
The post text is written when the post is ready and does not change after that. If a post has waited a long time, reread it before it goes out: the story may have new details by then.
Queue. A good post that does not need to go out right now can be queued with the "Queue" button in the bot, in the Publications block or in the story window. Queued posts go out one at a time: no more than once per chosen interval (once an hour by default), counted from the last publication in the project's channels and from posts already scheduled in Upcoming. If another publication went out close by, the queued post moves later, and only within the chosen hours (09:00 to 22:00 Moscow time by default). A post queued at night goes out in the morning. The queue is shown in My feed, in Publications → Queued: there you can pick another time, publish the post at once or remove it. If you open a queued post, you can edit the text and save it.
Many posts at once. Telegram limits how many messages a bot can send to one chat. If more than three posts are waiting for review at the same time, the bot sends one message "N posts are waiting for review" with a link to the site, where you can decide on them in the Publications block. If the bot message does not arrive, the posts still wait on the site.
A decision is counted once, wherever it is made: if a post was already published from the bot, the site shows "Already published", and the other way round.
#Publishing channels
First connect where to publish. Use the "Publishing channels" block, with Telegram, MAX and VK tabs.
#Telegram and MAX
- Add our bot to the channel or group. The bot links are right in the block.
- Channel: as an administrator with the right to post messages.
- Telegram group: a regular member is enough, as long as the bot is allowed to write.
- Enter
@channel, a link to it or its numeric id and click Add («Добавить»). - Click Check («Проверить»). The system makes sure the bot is there and can post. The status changes to "Connected".
If "Check" returns an error, it gives the reason: the channel was not found, the bot is not an administrator, the bot has no right to post, the bot is not in the group, or it is not allowed to write.
#Private group or channel without an @name
These have no @name address. Instead you need the numeric id, for example -1001234567890. You do not need to look it up by hand:
- Channel. Forward any post from the channel to the bot in a private message, and it replies with the channel id. The bot also sends the id privately to whoever added it to the channel, if you have messaged the bot before.
- Group. Forwarding does not help here: Telegram passes on the author of the message, not the group. But right after it is added, the bot posts a message with the id in the group. If that message is lost, send the
/idcommand in the group.
Paste the id into the channel field and click "Add", then "Check".
#MAX channel
MAX does not let a bot find a channel by an invite link max.ru/join/…, so the numeric id is the most reliable option. When you add the bot to a channel as an administrator, it sends you a private message with the channel name and id (a number like -79147257544831). Paste it and click "Check". A public link max.ru/name also works if the bot is already in the channel.
In MAX the post goes out with the same images and videos (up to 50 MB) as in Telegram.
#VK
Enter the community address (vk.com/club… or the short name) and the community access key. The key is created in community management, in the "API usage" («Работа с API») section. When creating it, grant access to the wall and photos: the post is published on the wall, and its images are first uploaded as community photos.
For now, only digests and posts published manually from the feed go to a community: online mode publishes to Telegram and MAX.
The key gives the right to post on behalf of the community. Do not share it with anyone except the service, and revoke it in VK settings if you disconnect the channel.
Each channel has its own switch: a disabled channel stays in the list, but posts do not go to it. How many channels you can connect depends on your plan: 3 on Pro, 5 on Max, across all projects.
#What to publish
By default, every story in the project feed that reaches the source threshold goes to the channel. The "What to publish" block narrows this down:
-
Only by monitoring keywords. Tick keywords, and only stories where at least one post matched one of them go to the channel. Next to each keyword you see how many times it matched this week. No AI model is involved here: the selection is exact and uses no tokens.
-
AI editor: story selection and post review (Pro and Max). Turn it on and describe in the "Channel topic" field what the channel is about and what should not get into it. The editor works twice:
- before the text, it compares the story with the description and answers "yes" or "no". Doubtful stories are skipped, and no post text is written for a rejected one;
- after the text, it rereads the finished post: is it a refusal instead of a post, is it about this story, does it contain facts that are not in the sources. If the post is over the limit, it shortens it instead of cutting it off mid-sentence.
Write selection criteria in "Channel topic", not in the post style: the style only controls how the text is written. The topic can be up to 1,000 characters.
The editor learns from your decisions. If "Show me the post before publishing" is on, the last 20 posts you sent or rejected yourself serve as examples for the editor: it sees what you usually take and decides closer to you. Posts that went out on their own after the deadline are not used as examples. Until there are decisions, the editor relies on the topic only.
The editor uses the token limit. Every check, before and after the text, uses the same monthly token limit as rewriting: about 200–300 tokens per story. The editor checks every story that passed the source threshold, so usage grows with the number of sources: 2,000 stories a day is about 15 million tokens a month. To lower usage, raise the source threshold or use the monitoring keyword filter: it removes stories before the editor and uses no tokens. "Test on recent stories" also uses the limit and shows how much.
Both filters can be on together: keywords are checked first, then the topic. The "Test on recent stories" («Проверить на последних сюжетах») button shows how the filters would decide on the last ten stories in the feed, before you save the settings.
A skipped story in the feed has the "Skipped" status with a reason: "no selected monitoring keyword", "off topic: …", "editor rejected the post: …", "cancelled manually". All decisions with the AI model's text are in the publishing log.
Do not write selection rules in the post style. "Publish only…", "reject…", "if it does not fit, answer SKIP": all of this is the job of the filters above. The AI model that writes the text starts making decisions by these rules itself and writes a refusal instead of a post. Such an answer will not go to the channel (the story gets "Skipped"), but the post will not go out either.
#Publishing mode
| Mode | What goes to the channel |
|---|---|
| Rewrite | The AI model rewrites the news from several sources into one post: in its own words, without repetition, in the style you set |
| Repost | The original post is forwarded unchanged: from Telegram to Telegram, from MAX to MAX |
Repost is available only when the story confirmation threshold is one source: you can forward a specific post, not a story made of ten. VK posts and website articles are skipped in Repost mode: there is no way to forward them to Telegram.
#When a post goes out
Three settings in minutes. All three are needed so that the post is complete but not late:
| Setting | What it does |
|---|---|
| Delay | How long to wait after the story forms |
| Extension | How much to postpone publication when someone else writes about the story |
| Max wait | The ceiling: the post goes out no later than this, however long sources keep writing |
The publication time is whichever comes later: "formation + delay" or "last new post + extension", but no later than "formation + max wait".
Delay 10, extension 5, maximum 30. The story forms at 12:00, so the plan is 12:10. At 12:08 another post arrives, and the plan moves to 12:13. Posts keep coming every few minutes and the plan keeps moving, but at 12:30 the post goes out anyway.
#Outdated stories
Online mode publishes only fresh news. There are two cases when a story does not go to the channel:
- It reached the threshold too late. If a story got the required number of sources more than 6 hours after it appeared, for online mode it is already yesterday's news. This happens when many sources are added to a project at once and their archive is loaded: old posts complete old stories.
- It sat in the queue. If a post did not go out within 3 hours of its planned time, for example because of a connection failure with the platform, the story is removed and marked "outdated". The publishing log gets one entry for all removed stories, with the "Queue" step.
The queue is shared by all projects, but each project gets its own share: a burst of stories in one project does not delay publications in others.
#Text settings
| Setting | What it does |
|---|---|
| Generation model | Which AI model writes the post. The senior model writes more accurately but is not available on all plans |
| Text limit | Maximum post length in characters. With the AI editor, a long post is shortened by meaning; without it, it is cut at the end of a sentence. The hint under the field reminds you of platform limits: a Telegram photo caption is 1,024 characters, a post without media 4,096, MAX 4,000 |
| Post style | How the post is written: a ready style ("News", "Channel post", "Flash", "Analysis") or your project's own style, on Pro and Max, created in project settings |
| Add source links | Add links to the posts the item was built from at the end of the post |
| Filter advertising | Skip stories that look like advertising. They get the "Skipped" status |
| Signature at the end of the post | A fixed line under every post, for example "Subscribe to @my_channel". A word can be made a link: the "Add link" button under the field inserts an entry like [Subscribe](https://t.me/my_channel), which appears in the post as the word "Subscribe" linking to the channel. A preview of the signature is shown under the field |
Remember to click "Save configuration" («Сохранить конфигурацию»). New delays immediately recalculate the time for stories already in the queue.
#How to write your own style
The AI model knows it is writing a news post; you do not need to explain that. What it needs is what makes your channel different from the one next door:
- Tone and audience. "We write for people in the region, no officialese, the way a neighbour would explain it."
- Structure. "First sentence: the point. Second: the consequences. No more than three paragraphs."
- Must have. "Always say exactly where it happened: district, street."
- Must not. "Do not add judgements or emotions that are not in the sources."
What should not be in a style is selection rules ("publish only…", "reject…"). The model that writes the text starts deciding by them itself and writes an explanation of why the story should not be published instead of a post. Selection is the job of the AI editor's "Channel topic". Your own style can be up to 3,000 characters.
Run autoposting into a test channel first and watch the posts for a day. A style almost always needs two or three edits, and it is better to make them before subscribers see the posts.
#Publishing log
At the bottom of the autoposting page is the log: what the service decided for each story.
- Published: which channels it went to and the full post text.
- Skipped: at which step and why: selection (did not match keywords, or the AI editor found it "off topic", with its explanation), advertising, token limit used up, safety check (the model wrote a refusal or nothing instead of a post; this text is shown too), the AI editor rejected the finished post, no channels.
- Error: the model did not answer, or no channel accepted the post, with the platform's response.
- Under review, Not publishing, Time ran out: the post waited for your decision: what was decided, by whom and where (in Telegram, MAX or on the site).
The buttons at the top filter the entries; the numbers on them are for 7 days. Click a row to expand the text.
#Scheduled
On the Scheduled tab, posts go out at set times rather than as soon as a story forms. This mode is available on Pro and Max plans.
Under "What goes out" («Что выходит»), choose a format:
- Top story: each release is one story as a separate post;
- Topic digest: a selection of posts over a period as one post, with a list and links.
Turn on the "Publish on schedule" («Публиковать по расписанию») switch and click Save. Under the buttons you see the time of the next release, and on the right the release log: what went out, which release was skipped and why. The same entries, with the "Scheduled release" step, are in the general publishing log.
Quiet hours apply to both formats: nothing goes out during these hours (Moscow time). A release that would fall into quiet hours moves to their end, and stories from the night become candidates for the first morning release.
#Top story
Each release takes one story: the biggest by number of independent sources among those updated since the previous release and that appeared no earlier than 6 hours before it. After you turn the mode on or resume it, counting starts from that moment, so stories that have not been updated for a long time do not resurface. If the top story does not pass the filters, the next biggest is taken. If there were no suitable stories, the release is skipped: no empty post goes to the channel, and the reason appears in the log ("no story reached the source threshold", "the AI editor rejected the candidates").
#When it goes out
The schedule is set on a 24-hour dial:
- At exact times: click the ring to add a release (the time is rounded to 15 minutes), click a mark to remove it.
- Every N minutes: from 30 to 360. The marks are placed automatically, counting from the end of quiet hours (or from midnight).
- Spread: up to ±30 minutes. Each release goes out at a random moment around its mark, so the channel does not look like it posts by alarm clock. On the dial this is the green shade around a mark.
- Days of the week: for example, weekdays only.
The centre of the dial shows how many releases there are per day; the grey arc is quiet hours. If you want more often, add releases: each release is one story, and several stories at once is already a digest.
#Top story post
The top story has its own post settings, separate from online mode: what to publish (all stories in the feed or only by monitoring keywords), the AI editor with the channel topic, style, text limit, source links, advertising filter and signature. The "Copy from online mode" («Взять из онлайн-режима») button copies the current settings of the Online tab; after that you can change them separately. Publishing channels are shared. There is no repost for the top story: it is a story from many sources, not one post by someone else.
While the top story is on, online publishing is paused; otherwise the channel would get both the stream and the releases. Stories that were waiting for online publication become candidates for the first release. Turn off the schedule or switch to the digest, and online mode continues with new stories.
#Topic digest
A digest is one post with a selection over a period: "Where to go this weekend in Pyatigorsk", "What's new in regional construction this week". The service finds suitable posts from your sources, picks the best and writes a list with links.
Open the Scheduled tab in Autoposting and choose "Topic digest" («Сводка по теме»). Publishing channels are the same as in online mode. A digest can run alongside online mode.
The digest schedule is by days of the week at a chosen hour, or every N hours from 1 to 168: for example, every 3 hours.
#How to set it up
- Digest topic. Describe in your own words what goes in and what does not: "things to do this weekend in Pyatigorsk: concerts, plays, exhibitions, festivals, events for children".
- Search queries. Click "Suggest search queries" («Подобрать поисковые запросы»), and the AI model suggests several queries on the topic. Remove an extra one with the cross, and add your own in the field below. Posts are searched by meaning, not by exact word match, so "events listing" also finds "join us for a concert".
- Title and number of items. The title goes into the post and onto the cover. From 3 to 12 items.
- Schedule. By days of the week at a set time (Moscow time) or every N hours (from 1 to 168). "Days in digest" is the period to collect posts from.
- Text style. "Digest (ready)" gives a date, a one-line intro and a list: what, where, when, and a "more" link. You can also choose your project's own prompt, the same ones as in online mode.
- Image. By default, a TeleMax-style cover with the digest title, date and number of items. You can upload your own image (JPEG, PNG or WebP up to 10 MB): with the "Title over the image" switch, the title is placed on it in our house style; without it, the image goes out as it is. A preview appears right under the button. The image goes to the channel together with the post.
Turn on the "Publish on schedule" switch and click Save. Under the buttons you will see the time of the next release.
#Check before publishing
"Build now, without publishing" («Собрать сейчас — без публикации») builds the digest and shows the post with its cover, but sends nothing to the channel. If you like the result, click "Publish this digest". Recent builds are saved, so you can see what went out and when.
If there are fewer than two suitable posts, the digest does not go out: no empty post goes to the channel.
#Where the content comes from
- Only posts from your project's sources over the chosen period. Add event listings, city communities and venue websites to the project, and the digest becomes fuller.
- The AI model drops past events and off-topic items when building the digest.
- On Pro and Max you can turn on the AI editor: it checks each item against the topic separately. The digest is cleaner but takes a little longer to build.
#Plan limits
Scheduled publishing, both digests and the top story, is available on Pro and Max plans. The digest period is up to 14 days; the AI editor and your own prompts are included.
The digest text and the top story post are written by an AI model, so each release uses rewrite tokens according to your plan.