By 2026, two of the most widely used content management systems for charities had official foundations for AI.
WordPress put an AI layer into its core software. Drupal has a coordinated AI initiative, with a 2026 roadmap backed by dozens of partner organisations. If your charity's website runs on either of them, AI features are now much easier to add, and suppliers are already pitching them.
This guide explains what's actually available on each platform, where AI genuinely helps charities, and what to get right before switching anything on. No hype, and no assumption that AI is the answer.
First: what does "AI in your website" actually mean?
When suppliers talk about AI in a CMS, they usually mean one of three things:
- Help for your team: drafting summaries, suggesting image descriptions (alt text), tagging content or rewriting in plain English, inside the editing screen your staff already use.
- Help for your visitors: a chatbot that answers common questions, or a search box that understands questions instead of just matching keywords.
- AI tools working with your site: letting AI assistants read or update your site on your behalf. Most platforms now do this through MCP (Model Context Protocol), a shared standard for connecting AI tools to other software, a bit like a universal plug.
Each platform below supports all three in different ways, and at different levels of maturity.
WordPress: AI built into the core
WordPress 7.0, released on 20 May 2026, added what the project calls "AI building blocks" to the core software:
- The WP AI Client: a single, shared way for plugins to talk to AI models. Instead of every plugin asking for its own AI account and settings, they can all use one connection.
- A Connectors screen: under Settings → Connectors, an administrator chooses which AI provider to use. Anthropic, Google and OpenAI are available by default.
- The Abilities API: introduced in WordPress 6.9, this lets the site describe what it can do, such as "create a draft post" or "list upcoming events", in a way AI tools can understand.
- The MCP Adapter: a separate, official plugin that lets external AI assistants use those abilities. Each ability has to be deliberately made available.
What it means for charities: WordPress doesn't come with an AI provider switched on. Nothing is sent to an AI service unless someone configures a provider and a plugin uses it. You choose the provider, you control the account, and you can change your mind later without rebuilding the plugins that use it. That's a sensible, cautious design, and a good foundation for AI features that are easy to switch off.
Drupal: an ambitious AI roadmap
Drupal has taken a broad, community-led approach. Its AI module ecosystem is vendor-agnostic, so it isn't tied to one AI company. It includes:
- Editorial tools: AI help inside the content editor, such as drafting, summarising and suggesting alt text.
- Chatbots and search: modules for visitor-facing chat and AI-powered search, though some, like AI Search, were still in early testing as of September 2026. The AI project page lists the current related modules.
- Agents: the AI Agents module can help set up site structures such as content types and fields, and the MCP Server module lets external AI tools work with a Drupal site.
- Management: the AI Dashboard gives administrators one place to manage AI features and providers.
Drupal's 2026 roadmap goes further, aiming for "background agents" that work on schedules and triggers within editorial approval workflows. Independent reviewers describe the editorial features as ready for production use, with the more autonomous agent features still maturing.
A note from us: Bee Lite doesn't build on Drupal, so we're describing the landscape rather than recommending a setup. If your site runs on Drupal, talk to your Drupal supplier about which of these features are stable enough for your needs.
At a glance
| Platform | What's official | Best suited to |
|---|---|---|
| WordPress | AI Client and Connectors in core (7.0), Abilities API, MCP Adapter plugin | Websites wanting simple, switchable AI features through plugins |
| Drupal | AI module ecosystem, agents, MCP, 2026 AI roadmap | Large, structured content sites with an experienced Drupal team |
Where AI genuinely helps charities
The best uses we see are modest, specific and checked by a person:
- Accessibility support: drafting alt text for images, which an editor then checks. It's faster than starting from scratch, but AI descriptions can miss the point of an image. See our accessibility guide.
- Plain-English rewrites: turning dense policy or research content into clearer summaries for the public.
- Helping people find things: search or a chatbot that answers "how do I volunteer?" or "do you cover my area?" from your own published content, and points to the right page.
- Sorting enquiries: tagging incoming messages by topic or urgency so the right person sees them sooner.
- Content housekeeping: suggesting tags, spotting outdated pages and summarising long reports for newsletters.
What to get right first
Safeguarding
If people in distress might use your website, a chatbot must never be the only response. It should recognise when to stop, signpost to human help and crisis lines, and never give advice it isn't qualified to give. Test it with the hardest questions your service users might ask, not just the easy ones.
Data protection
Be careful about what information reaches an AI provider, especially anything about service users, health or finances. The ICO expects a data protection impact assessment (DPIA) for most uses of AI that involve personal data. Check each provider's terms on data retention and whether your data is used to train their models.
Accuracy
AI can be confidently wrong. Anything published, or anything that gives people guidance, needs human review, or needs to be limited to answering from your own approved content.
Transparency
Tell people when they're talking to AI, and make it easy to reach a human. Trust is your charity's most valuable asset, and it's easy to lose.
Accessibility and cost
Chat widgets are often hard to use with a keyboard or screen reader, so test them like any other feature. AI services are usually charged by usage, so set spending limits and watch costs as use grows.
How we approach AI projects at Bee Lite
- Problem first, tool second. We start with the job that needs doing. Sometimes the answer is AI. Sometimes it's a clearer page or a better form.
- Start small. One focused pilot, such as an FAQ assistant or alt text suggestions, measured against a clear goal before anything grows.
- Stay provider-agnostic. We build on WordPress's official AI foundations so you can switch AI provider without starting again.
- Send as little data as possible. AI features only see the information they need, and sensitive data stays out unless there's a clear, assessed reason.
- Keep people in charge. Human review for anything published, clear routes to a real person, and an easy off switch.
Find out more about our AI workflows service, including customised chatbots.
Five questions to ask before adding AI
- What specific problem will this solve, and how will we know it worked?
- What information will the AI see, and does any of it need a DPIA?
- Who checks its output before it reaches the public?
- What happens when it gets something wrong, or someone needs a human?
- What will it cost each month as use grows, and can we switch it off easily?
Frequently asked questions
Do we need to change platforms to use AI?
No. WordPress and Drupal both now have official AI foundations. The better question is whether your current site is up to date enough to use them safely. Our guide to unsupported software can help you check.
Will AI write our content for us?
It can draft, summarise and suggest. But your charity's voice, accuracy and judgement have to come from people who understand your cause, and supporters can usually tell the difference.
Is our data used to train AI models?
It depends on the provider and the type of account. Many business and developer services don't train on your data by default, but you should always check the terms before connecting anything.
Is AI expensive?
A small, focused feature can cost very little to run. Costs grow with usage, so set spending limits and review them monthly.
Useful, not magical
AI is now a standard part of the platforms most charities use. Used well, it saves staff time, makes information easier to find and supports accessibility. Used carelessly, it risks trust, data and safeguarding. The difference is in the planning, not the technology.
Sources
- WordPress 7.0 Field Guide (make.wordpress.org)
- WP AI Client and MCP Adapter repositories (GitHub)
- Introducing the Connectors API in WordPress 7.0 (make.wordpress.org)
- Drupal's AI roadmap for 2026 (Dries Buytaert)
- Drupal CMS product strategy (drupal.org)
- Guidance on AI and data protection (ICO)

