At least 826,000 WordPress sites run a chat widget, and almost none of it is AI. That number comes from querying the WordPress.org plugin API directly, not from a listicle, because the listicles for this category are all affiliate pages. The shape of the number changed what we decided to build.
Here is the scan, the judgement calls inside it, and what it does not tell you.
The method, so you can repeat it
The WordPress.org plugin API is public and takes no key. Six searches, one hundred results each, deduplicated by slug, filtered to plugins with at least a thousand active installs:
https://api.wordpress.org/plugins/info/1.2/
?action=query_plugins
&request[search]=ai chatbot
&request[per_page]=100
&request[fields][active_installs]=1
&request[fields][ratings]=1Search terms: ai chatbot, chatgpt, ai chat, chatbot, ai assistant, live chat ai.
That produced 175 plugins over the install threshold, of which 38 are genuinely chat or AI
chat rather than a spam filter or an SEO plugin that happened to match on the word “AI”.
One caveat worth stating: WordPress reports active installs in buckets, not exact figures. A plugin showing 100,000 is somewhere between 100,000 and 199,999. Every number below is therefore a floor, not a point estimate.
Finding one: the market is enormous and it is not AI
Those 38 plugins cover at least 826,000 active installs. That is a lot of WordPress sites that have already decided they want a chat bubble in the corner.
But look at what is actually at the top:
| Plugin | Installs | Rating | Is it AI? |
|---|---|---|---|
| Chaty | 300,000 | 5.0 | No. It is a button that opens WhatsApp |
| AI Engine | 100,000 | 4.9 | Yes |
| Tawk.to | 100,000 | 4.6 | No. Human live chat |
| Tidio | 70,000 | 4.7 | Partly |
| Chatway | 30,000 | 5.0 | No. Human live chat |
The single most-installed “chat” plugin on WordPress, by a factor of three, is a floating button that opens WhatsApp. It does not answer anything. It forwards you to a human.
That is the real shape of this market. Several hundred thousand site owners have decided that visitors have questions worth capturing, and the overwhelming majority of them are still solving it by routing the question to a person.
Finding two: there is one serious AI incumbent, and it is very good
AI Engine sits at 100,000 installs with a 4.9 rating across 864 reviews. Only 1% of those reviews are one or two star. It shipped an update the day I ran this query.
A 1% bad-review rate across 864 reviews is not luck. For comparison, Tawk.to is at 9% and Tidio at 6%, and both are perfectly respectable products. AI Engine is genuinely loved by the people using it.
If you are considering building here, that number is the one to sit with. Competing on “free AI chatbot for WordPress” against a free, mature, beloved incumbent with a five-year head start is not a strategy. It is a way to spend a year discovering something you could have learned from one API call.
Finding three: the gap is not features, it is finish
Read AI Engine’s own description and it tells you exactly what it is: “The Chatbot, AI Framework & MCP for WordPress”. A framework. To get value out of it you bring an API key, choose a model, build the knowledge base, wire up the embeddings, and style the widget.
That is a completely reasonable thing to be, and for developers it is the right tool. But “framework” and “product” are different jobs, and most people running a shop want the second one.
There is a second gap, and it is the one nobody in the category seems to notice. Install forty of these plugins and look at the output. You get forty variations on the same rounded bubble, most of them purple or blue, most with a cartoon robot avatar. A chatbot is the only plugin whose output is visible to every visitor on every page of the site, and the entire category treats it as plumbing.
Finding four: 5 of the 38 are effectively abandoned
Five plugins in the set have not shipped an update in over six months. One of the top twenty, at 20,000 installs, was last touched in April 2025. Another, at 100,000 installs, in January 2026.
For a plugin category that talks to an AI provider’s API, a year without an update is not neutral. Model names get deprecated, endpoints change, and the plugin quietly stops working on somebody’s site.
When is this scan the wrong evidence?
Four times, and they are worth saying before anyone quotes the 826,000 back at me.
An active install is not a user. It counts sites with the plugin enabled, not sites whose visitors ever open the chat. A plugin installed during a redesign and forgotten about counts exactly the same as one answering fifty questions a day.
The directory only holds the free ones. WordPress.org lists plugins distributed through WordPress.org. Anything sold from a vendor’s own site does not appear here at all, so the paid end of this category is invisible to this method. The scan describes the free market and should not be quoted as describing the whole one.
Six search terms is a judgement, not a census. A plugin that describes itself without using the words I searched for is simply absent from the set. So is any plugin under a thousand installs, which is where new entrants live.
The 175 to 38 cut was made by a person. Deciding that an SEO plugin matching on the word “AI” is not a chat plugin is obvious; deciding where a helpdesk with a chat module belongs is not. Someone repeating this with the same API calls and a different threshold would get a different number, and should.
None of that moves the finding. The top of the market being a WhatsApp button survives every one of these caveats, because it is a comparison inside the same flawed measure.
What we concluded
Three things, and they are now the plan:
- Do not compete on free. AI Engine owns that and deserves to.
- Compete on finish and on design. The knowledge base should build itself. The default widget should look like a studio made it, because it is the one thing every visitor sees.
- Say plainly where the free option wins. Our comparison page names it: if you want to build your own AI features, AI Engine is the better foundation. Pretending otherwise just makes the rest of the page less believable.
The whole scan took about four minutes of API calls. It is worth doing before you build anything in a category you did not invent.
What owners ask us most
How many WordPress sites have a chat plugin installed?
At least 826,000, counting only plugins with a thousand or more active installs that are genuinely chat or AI chat. WordPress reports installs in buckets rather than exact figures, so a plugin showing 100,000 sits somewhere between 100,000 and 199,999. Every number here is a floor rather than a point estimate, and the real total is higher, probably by a wide margin.
What is the most installed chat plugin on WordPress?
Chaty, at 300,000 active installs, and it is not AI at all. It is a floating button that opens WhatsApp or another messenger. It answers nothing and forwards the visitor to a human, which tells you how most site owners still solve this problem. Three times more sites reach for that than for the largest AI option in the set.
Is AI Engine the main competitor for a WordPress AI chatbot?
Yes. It sits at 100,000 active installs with a 4.9 rating across 864 reviews, only 1% of which are one or two star, and it ships updates constantly. Competing with it on price would be a mistake, because it is free, mature and genuinely well liked. For comparison, Tawk.to runs at 9% bad reviews and Tidio at 6%, and both are respectable products.
How do I run this market scan myself?
Query the WordPress.org plugin API directly. It is public, needs no key, and accepts a search term plus the active_installs and ratings fields. Six searches at a hundred results each, deduplicated by slug, took about four minutes and needed no paid tool. The judgement is not in the querying, it is in deciding which 38 of the 175 results are genuinely chat.
Why does the design of a chatbot plugin matter commercially?
Because a chatbot is the only plugin whose output appears to every visitor on every page. Install forty of them and you get forty versions of the same rounded bubble, most of them purple or blue, most with a cartoon robot avatar. It is a design surface that the entire category treats as plumbing, which makes it the cheapest place to be visibly different.
The full comparison against the incumbent is on the AI Engine comparison page, the positioning that came out of this is visible on the features page, the engineering side of the same “measure it first” habit is in why we dropped GSAP, and what the category actually contains is in the guide to AI chatbots for websites.
