Writing for Algorithms

The Core Principle

Algorithms are trained to detect two things: spam and low-quality content. If your writing doesn't look like spam and doesn't read like filler, you're already ahead of most content on the internet. The goal isn't to trick algorithms. It's to write clearly and avoid patterns that trigger suppression.

X (formerly Twitter)

X's algorithm rewards engagement and penalizes behavior that looks automated or promotional.

What Works

• Conversational tone. Write like you're talking to a person, not broadcasting. Tweets that sound natural get more replies, and replies boost your reach.
• 1-2 hashtags maximum. More than two hashtags makes your tweet look like spam. X's algorithm actively suppresses hashtag-stuffed posts.
• Context for links. Don't just drop a URL. Add a sentence explaining what it is and why it matters. Bare links get less distribution.
• Threads over single tweets for long-form. Threads get algorithmic boosts when people read through multiple posts.

What Gets Suppressed

• Identical text posted multiple times.
• Aggressive call-to-action language ("FOLLOW ME," "RETWEET THIS").
• External links without context (especially to sites X considers competitors).
• Rapid-fire posting - more than 5-10 tweets in a short window.

Medium

Medium's recommendation engine favors long-form, original content. The algorithm cares about read time, completion rate, and reader engagement.

What Works

• Depth over clickbait. Articles that deliver on their headline promise get recommended. Articles that bait readers and underdeliver get buried.
• Original research and data. Medium's algorithm and curators favor content with unique insights. Rephrasing what everyone else says won't cut it.
• Clear structure. Use headings, short paragraphs, and lists. Medium tracks how far readers scroll. Good structure keeps them reading.
• Strong opening paragraphs. You have about 10 seconds to hook a reader. If they leave immediately, the algorithm notices.

What Gets Suppressed

• Thin content under 800 words that doesn't provide real value.
• Clickbait headlines that don't match the content.
• Excessive self-promotion or affiliate links.
• Content that's clearly republished from elsewhere without meaningful additions.

Facebook

Facebook has become the hardest platform for organic reach. The algorithm heavily favors native content and community interaction.

What Works

• Native content over links. Posts with external links get dramatically less reach than text-only or image posts. If you must share a link, put it in the comments.
• The 80/20 community ratio. For every promotional post, share four pieces of value - tips, questions, stories, or community highlights. Pages that only promote get throttled.
• Questions and discussion starters. Facebook's algorithm boosts posts that generate comments. Ask genuine questions that your audience wants to answer.
• Personal stories. Facebook prioritizes content that feels personal and authentic over corporate-sounding updates.

What Gets Suppressed

• Engagement bait ("Like if you agree," "Tag a friend who..."). Facebook explicitly penalizes this.
• Posts with multiple external links.
• Overly polished, ad-like content from pages.
• Text heavy with ALL CAPS or excessive emoji.

Avoid Spam Patterns

Every platform's algorithm is trained on millions of spam examples. The patterns are consistent:

• Financial promises: "earn $X," "make money fast," "guaranteed income."
• Urgency language: "act now," "limited time," "don't miss out."
• Excessive formatting: ALL CAPS, multiple exclamation marks, emoji spam.
• Vague calls to action: "click here," "check this out," "you won't believe."

TextScore flags all of these. Run your content through the spam check before publishing on any platform.

Keep Sentiment Balanced

Extreme sentiment - either very positive or very negative - can trigger algorithmic reviews. Content that reads as overly promotional (extremely positive) or hostile (extremely negative) gets flagged.

Aim for a natural sentiment range. Honest, balanced content performs better than hype or outrage in the long run. TextScore's sentiment analysis shows you where your content falls on this spectrum.

Write in Natural Language

Algorithms are getting better at detecting unnatural writing. This includes:

• Keyword stuffing - repeating the same phrase unnaturally.
• Robotic sentence structures - every sentence following the same pattern.
• Jargon overload - using industry terms when plain language works.
• AI-generated text that follows predictable patterns.

Write the way you'd explain something to a smart friend. That's the register most algorithms reward.

Before publishing on any platform, run your content through TextScore and check these scores:

• Spam score: Keep it below 15. Anything higher means you have patterns that algorithms will flag.
• Readability: Match your platform's sweet spot. 70+ for social, 55+ for blogs, 45+ for technical content.
• Sentiment: Stay in the balanced range. Extreme scores in either direction are a warning sign.
• Link safety: Every link should pass. One bad link can tank an otherwise clean post.