Spam used to be easy to spot.
In the early CrazyEngineers days, a spammer would sign up, make a post or two with a link, and disappear. The smarter ones came back weeks later and quietly edited a link into an old post, hoping no moderator would notice. They wanted the link to stay for the SEO value, and hoped a few members would click.
We were pretty hard on moderation. Pure advertisers got banned the same day. It was annoying, but it was simple.
That world is gone.
Today, a lot of spam comes from real people, hired in large numbers to spam forums and communities. They sit at real computers, so ordinary spam protection doesn't catch them. They sign up like anyone else. They use AI to write a few random, useful-looking posts to build trust. And then, once nobody's watching closely, they start adding links.
Very difficult to catch.
I've been running online communities for twenty years. CrazyEngineers has 400,000+ members, and today I build Jatra, a community platform, where we built our spam defences around exactly this problem. Here's how I'd stop forum spam today.
Why spam is harder now
Two things changed at the same time.
First, bots got better. Imperva's 2025 Bad Bot Report found that automated traffic made up 51% of all web traffic in 2024, overtaking humans for the first time in a decade, and that bad bots alone were 37%. AI made it cheap to build bots that look human.
Second, and this is the part that hurts communities most, humans got AI. A paid spammer used to write one clumsy post at a time. Now they can write ten fluent ones in a minute. The post looks fine. The grammar is fine. It even answers the question, sort of.
So the old tricks don't work anymore. A CAPTCHA stops a script, not a person. A banned-words list doesn't stop a polite, well-written post. You need a different approach.
Why it matters for a business community
Spam isn't just ugly. It costs you.
Your members leave. Nobody wants to wade through link drops to find a real answer, and the people who leave first are your best contributors.
It also attracts more spam. Spammers want their links in search engines, and Google says as much in its guide to preventing user-generated spam. A community where spam stays up is a community that gets targeted harder.
And it puts your brand next to whatever they're selling. On your own domain.
Layer 1: Make spam worthless
The easiest spammer to stop is the one who gets nothing for their effort.
Most community spam exists for the link. So take away the value of the link. Add rel="ugc nofollow" to links posted by members, so search engines don't pass any credit through them. Google recommends exactly this for untrusted content.
On Jatra, this happens automatically. Outside links in members' posts get ugc nofollow. Links posted by owners and staff stay normal, because you trust them.
Also: drop signatures. They're a legacy forum feature, from the vBulletin and XenForo days, and frankly speaking, they're the place meant for spam. Jatra doesn't have them.
Layer 2: Slow down new accounts
Spammers are impatient. Make them wait.
New members don't need full powers on day one. Give them a trust window: a few weeks where their posts get a little more attention before they're treated like everyone else.
On Jatra, every new member has a trust window, 30 days by default (you can set anywhere from 1 to 90). Until they're past it, and have at least three published posts, their public posts are screened quietly in the background. Members don't see a restriction. Their posts just get checked.
You can go further. Owners can switch on an option that stops members in their trust window from adding links and images at all. It's off by default, and members can see when it applies. It leaves new people no option but to contribute first.
This is also what caught one of our trickiest cases: a member who wrote genuinely useful posts for about two weeks, then posted spam links overnight on day 17. Because they were still inside their first 30 days, the links were caught before anyone saw them. I wrote about it in how to manage an online community.
Layer 3: Let AI read the first posts
Here's the irony. The best tool against AI-written spam is AI.
A human moderator can't read every post from every new member, especially in a busy community. But AI can, instantly, at the moment someone hits publish.
That's what spam screening in Jatra AI does. When a newer member publishes in a public channel, AI reads the post straight away, day or night, and scores it for spam and unrelated promotional links. Anything likely to be spam is held for a moderator to approve or reject, so nobody has to watch the feed overnight. It doesn't delete anything on its own. A human always makes the call.
How good is it? On CrazyEngineers, it flags about 8 to 15 posts a week, and 90 to 95% of what it flags turns out to be real spam. Some posts get flagged by mistake, and that's okay. A moderator approves them in a click, and the system gets better with time. Today it reads text only, not images.
It's included on every Jatra plan, with no extra charge.
Layer 4: Keep humans for the judgement calls
AI is great at the obvious stuff. It can't judge intent.
Sometimes a well-meaning member promotes their product too much. Sometimes a genuine expert drops a link that's actually the best answer. Sometimes a post sits right on the line.
That's human work. Let trusted members report posts. Give moderators a simple queue. And use one rule: promote your product, but first create value. Someone who's helped a dozen people can mention what they've built. Someone whose first post is a link can't.
How to spot the patient spammer
The new breed of spammer is patient, so look for patterns, not single posts.
- Posts that sound right but say nothing. Generic, fluent, agreeable. They summarise the question back, add a few safe points, and never share an actual experience.
- A sudden change in behaviour. Weeks of harmless posts, then links. Often overnight, when moderators are asleep.
- Edits to old posts. A link quietly added to something written weeks ago. This one is as old as forums.
- Profiles with nothing behind them. No introduction, no replies to other people, no back-and-forth. Just posts.
- Answers that don't match the community. Perfect English in a community where everyone writes casually. Topics that drift slightly off your niche.
None of these alone proves anything. Together, they're worth a closer look.
When you catch one, clean up fast
Remove everything the spammer posted, not just the post you noticed. Ban the account. And check for edits to their older posts, because that's where the links hide.
Then look at how they got in. Did they post links on day one? Switch on the link option for new members. Did they slip through the trust window? Maybe lengthen it. Every spammer teaches you something.
Your spam defence checklist
Swipe sideways to see every column.
| Layer | What it stops | On Jatra |
|---|---|---|
ugc nofollow on member links |
Spammers who want SEO value | Automatic for members; staff links followed |
| No signatures | Link drops hidden under every post | Not offered, on purpose |
| Trust window | Brand-new accounts posting freely | 30 days by default (1 to 90), plus three published posts |
| No links or images for new members | Day-one link drops and image spam | Optional, off by default |
| AI screening of first posts | Fluent, AI-written spam from newer members | Text in public channels, day or night, held for a human to review |
| Member reports and a review queue | Everything the machines miss | Report button and review queue on all plans |
Frequently asked questions
How do you stop spam on a forum?
Use layers. Make member links ugc nofollow so spam gets no SEO value, slow down new accounts with a trust window, screen new members' first posts, and let trusted members report what slips through. No single tool catches everything.
Do CAPTCHAs stop forum spam?
They stop automated scripts, which still helps. They don't stop paid humans or people using AI to write posts, and that's where most of today's community spam comes from.
Can AI detect AI-written spam?
It can catch most of it, especially when it looks at new members' posts at the moment they're published. It isn't perfect, so the best setup holds suspicious posts for a human to approve rather than deleting them automatically.
Should you ban AI-written posts?
Not necessarily. Plenty of genuine members use AI to tidy up their writing. Judge the post, not the tool: is it useful, and is it written for the community or for a link?
What does rel="ugc" mean?
It's a link attribute that tells search engines the link was posted by a user, not by the site owner. Combined with nofollow, it means the link passes no ranking credit, which takes away a big reason to spam.
Where I'd start
If I were fighting spam in a community tomorrow:
- Make sure member links are
ugc nofollow. - Turn off signatures.
- Set a trust window for new members.
- Screen first posts, with AI if you can, and a human approving.
- Let members report posts.
- When you catch a spammer, remove everything and check their edits.
That's it. You won't stop every spammer. But you can make your community the most boring target on the internet, and they'll move on.
For the rest of the daily work, read how to manage an online community. If spam is eating your moderators' time, book a call and tell us what you're seeing.