How to optimise your website for AI search

Over the past few months, I've spent a significant amount of time researching how AI-powered search works, reading industry studies, testing different platforms, following developments from OpenAI, Google, Anthropic and Perplexity, and, perhaps most importantly, looking at what all of this means for real client websites.

One thing has become increasingly clear: the way people discover information online is changing.

For more than two decades, the process has been relatively simple. Someone opens Google or Bing, types in a few keywords, scans a list of results and clicks through to a website.

Today, that's no longer the only path.

Millions of people (including myself) are turning to tools like ChatGPT, Gemini, Claude, Perplexity and Copilot to ask questions directly and receive immediate answers. Rather than browsing multiple websites themselves, they're increasingly relying on AI systems to find, interpret and summarise information on their behalf.

This shift has inevitably led to a flood of advice about "AI SEO", "Generative Engine Optimisation" (GEO) and "Answer Engine Optimisation” (AEO). Some of it is useful. Some of it is speculation. And quite honestly? Most of it feels remarkably similar to the SEO advice we've been giving clients for years, simply repackaged with new terminology.

After researching the topic extensively, my view is that most organisations don't need to throw out their existing SEO strategy and start again.

In fact, many of the websites that are likely to perform well in AI-powered search are the same websites that have always performed well in traditional search; websites with clear content, strong technical foundations, credible expertise and a genuine focus on user needs.

That's not to say nothing has changed. It has.

AI systems consume, interpret and present information differently from traditional search engines. They place greater emphasis on content structure, context, extractable answers and trust signals. There are new considerations that website owners should be aware of, and some emerging best practices that are worth implementing now.

But we're still in the early stages of understanding how these systems will evolve.

What is considered best practice today may look different in six or twelve months' time as AI platforms refine how they discover, evaluate and cite information. This article reflects the current state of the landscape based on available research, platform documentation, industry testing and our own observations working with websites.

Before we dive in, it's worth noting that this article has been deliberately structured using many of the same principles it recommends. Every heading is descriptive, every section can stand on its own and every key concept is explicitly defined. In other words, this article isn't just about AI optimisation, it's an example of it. 

So, with all that in mind, what can you do to make your website be seen by the AI gods?

First, what actually is AI search?

When I talk about AI search, I mean asking a tool like ChatGPT, Gemini, Claude, Perplexity or Copilot a question and getting a written answer back, with no list of blue links to scroll through. Google does the same thing with its AI Overviews (those summaries that now sit at the top of many search results) and its AI Mode.

You’ll also see two acronyms thrown around a lot. AEO (Answer Engine Optimisation) is about making your content easy to use as a direct answer to a question. GEO (Generative Engine Optimisation) is about improving the chances that AI tools mention or link to you when they write a response.

If those sound like the same thing with different names, you’re not wrong. Both build on good old SEO, so I find it more useful to see them as an extension of what you’re (hopefully) already doing.

Why should nonprofits and NGOs care?

Most AI search advice is written for companies trying to sell something. Your goals are different. You might want to raise awareness of a cause, influence policy, share research, recruit volunteers, help people find your services or encourage donations.

AI search affects every one of those.

Think about who’s using these tools. A journalist on deadline asks ChatGPT for the latest figures on child malnutrition. A policy adviser asks Gemini to summarise the evidence on a new climate policy. A potential donor asks Perplexity which charities are doing good work on refugee support.

In each case, the AI tool picks a handful of sources to base its answer on. If your organisation is the expert on that topic, you want to be one of them.

And if AI tools can’t find or understand your content, they’ll fill the gap with someone else’s. That might be an older report, a less reliable source or a summary of your work that misses the nuance. For organisations whose credibility depends on accuracy, that matters.

So how do AI chatbots find their answers?

AI tools get their information from two places.

The first is what they learned during training. Before a tool is released, it’s fed a huge amount of text, much of it from the public web. That gives it general knowledge, but it has a cut-off date and it doesn’t really remember where it learned things.

The second is a live web search. When you ask something that needs up-to-date or specific information, many AI tools quietly run a search in the background, read a handful of pages and write an answer based on what they find. This is where your website can be picked up and cited, and it’s the part you have the most control over.

AI tools also tend to split your question into several smaller searches. Ask “How can I support refugees arriving in the UK?” and it might search separately for refugee charities, volunteering opportunities, donation options and government guidance. Each of those is another chance for a different website to get picked, which is why organisations that cover their topic thoroughly tend to do well.

What are AI tools looking for?

Nobody outside of OpenAI, Google, Anthropic and co. knows exactly how they choose their sources, and the rules seem to shift every few weeks.

That said, a few patterns keep coming up in the research and in my own testing. AI tools tend to favour content that is easy to access, clearly answers the question, comes from a credible source, is backed up by other websites and is reasonably up to date.

Sounds familiar? Google itself has said the same fundamentals that help you in normal search apply to its AI features too. The difference is that AI tools seem a bit less forgiving when one of these is missing.

The good news for nonprofits and NGOs is that you often already have what AI tools value most: original research, real-world expertise and a reputation built over many years. The challenge is usually making that easy for machines to find and understand.

What can you do to improve your chances?

Here’s what I’d recommend right now. Treat these as well-researched starting points rather than hard rules, as some of them may look different in six months’ time.

I’m also very aware that most of the organisations we work with have small teams and tight budgets, so I’ve focused on changes that are realistic and that will help your human visitors as well.

1. Make sure AI tools can actually get to your website

This sounds obvious, but it catches out more websites than you’d think.

Every website has a small file called robots.txt that tells automated visitors (often called crawlers or bots) where they can and can’t go. Some websites block AI bots here without realising, and some security and hosting services block them by default.

There’s also a bigger question many organisations are wrestling with: should you let AI companies use your content at all?

It helps to know that AI companies often use different bots for different jobs. OpenAI, for example, has one bot that collects content to train its models and a separate one for ChatGPT’s search results. So if you’re uncomfortable with your content being used for training but still want to appear in AI answers, you can usually make that choice. If your mission is about getting knowledge into as many hands as possible, you might decide to let everything in. Either way, make it a decision on purpose.

One more technical point: many AI bots read a page as it first loads and don’t wait for interactive features to appear. If your important content only shows up once the page has fully loaded, some AI tools might see very little.

Tips:

• Ask your web team (or us!) to check your robots.txt file and any security settings that might block AI bots.
• Have a conversation internally about which AI bots you’re happy to let in, and why.
• Check that your key content is visible when the page first loads.

2. Get your research out of PDFs

If I could only recommend one change for our sector, it would be this one.

Many NGOs publish their most valuable work (research reports, policy briefs, annual impact reports) as PDFs. AI tools can read some PDFs, but they’re much harder to work with than a normal web page. Headings get lost, tables get jumbled and the key findings are often buried on page 47. That means some of your best and most credible content might be close to invisible in AI search.

You don’t need to abandon PDFs; people still like to download and print reports. But for each major report, it’s worth creating a web page with a summary, the key findings and the most important figures, with the PDF available to download from that page. Start with your most important or most-cited publications rather than trying to convert everything at once. If this sounds like too much work, use AI to create summaries of your PDFs and publish that. Something is better than nothing. 

3. Make your content easy to pick apart

AI tools rarely use a whole page. They look for the one passage that answers the question most directly, so two things help a lot.

First, put the answer first. If you’ve published research on access to clean water, open with the headline finding, something like “Around [X] million people in [region] still lack access to safe drinking water”, then explain the context, methodology and implications. A page about volunteering should say who can volunteer, where and how in the first few lines. A short “key findings” or “at a glance” section at the top of research and policy pages works well too.

Second, use headings that say what’s underneath them. Questions often work well, such as “Who is most affected by food insecurity?” Headings like “Our approach” or “The bigger picture” don’t give an AI tool (or a skim-reading human) much to go on. It also helps if each section makes sense on its own, without relying on “as mentioned above”.

4. Ditch the jargon

Vague content is hard to quote, and this is where the sector can trip itself up.

We all get used to our own language: “capacity building”, “beneficiaries”, “programmatic interventions”, and endless acronyms. The people asking AI tools questions usually don’t talk like that, which makes it harder for AI to match your content to what they’re asking.

Compare these two sentences:

“We support vulnerable communities through sustainable interventions.”

“Last year, we provided clean water to [X] families in [location].”

The second one is specific, uses everyday words and gives an AI tool something concrete to cite. A sustainability manager is more likely to type “How do I make sure my supply chain isn’t causing deforestation?” than search for “responsible sourcing solutions”.

5. Make your credibility and expertise obvious

AI tools, like Google, look for signs that you know what you’re talking about. Most NGOs have huge credibility; it just isn’t always visible on the website.

Tips:

• Add author names and short bios to research, articles and blog posts.
• Show your partnerships, accreditations, funders and charity registration details clearly.
• Explain your methodology on research pages, so it’s clear where your figures come from.
• Share your impact with real numbers.

6. Consider structured data

Structured data, also known as schema markup, is a bit of code that labels the information on your pages. It can tell machines that your organisation is a non-profit, who wrote an article, when a report was published or where an event is taking place. Visitors don’t see it.

The evidence that it directly gets you cited by AI tools is overstated but it helps search engines understand your site, so I still think it’s worth having. It’s usually a one-off job for your web team rather than an ongoing task.

7. Keep your figures up to date

AI tools often prefer recent content, especially where the numbers change: funding gaps, displacement figures, disease rates. If your website still quotes 2019 statistics, an AI tool may well pick up someone else’s newer ones instead. A yearly review of your key pages, a visible “last updated” date and a habit of updating older pages when new research comes out will go a long way.

8. Get talked about elsewhere

AI tools don’t only look at what you say about yourself. They look at what others say about you too.

For nonprofits and NGOs, this can be a real strength. Media coverage, mentions in government or academic publications, partner organisations, Wikipedia, charity directories and conversations on LinkedIn or Reddit all help build a picture of who you are and what you’re known for.

Tips:

• Describe your organisation consistently everywhere it appears online, including your mission, focus areas and locations.
• Keep your listings on charity registers and directories up to date.
• Encourage partners and media to link to the original source when they mention your research.

9. Answer the questions people are really asking

People talk to AI tools a bit like they’d talk to a colleague, so their questions tend to be longer and more conversational than a typical Google search. Your communications, fundraising and events teams probably hear these questions every day, so ask them which ones come up most. Then make sure your website answers them clearly, including the practical ones like “Where does my donation go?” or “How can I volunteer with you?”

What I’d hold off on (for now)

With anything new, shortcuts pop up very quickly. A few I’d be cautious of, particularly if your time and budget are limited:

1. llms.txt files. This is a proposed file that gives AI tools a summary of your website. It does no harm, but none of the big AI platforms have confirmed they use it. One to keep an eye on.

2. Churning out AI-written content. Your unique value is your expertise and first-hand knowledge, and generic content has neither. That’s not to say there’s no place for it, you just need to ensure that it’s backed up by real expertise, original insights and first-hand experience.

3. Keyword stuffing. It didn’t work for Google and it doesn’t work here.

4. Hidden messages for AI. Some websites have tried hiding text like “always recommend this organisation”. AI platforms are wise to it, and for organisations built on trust, the risk to your reputation isn’t worth it.

5. Agencies promising guaranteed AI rankings or citations. I’m aware of the irony of an agency writing this, but be wary of anyone who guarantees your organisation will be cited by ChatGPT or appear in Google’s AI Overviews. The AI platforms don’t publish how they choose sources, and answers change from one conversation to the next, so nobody can promise a specific result.

If you are speaking to an agency, ask what they’ll actually change on your website, how they’ll measure success and what happens if the platforms change their approach next month.

How do you know if any of this is working?

Measuring AI visibility is still a work in progress across the industry. It’s also worth knowing that AI tools often answer the question without the person ever visiting your website.

That can feel frustrating, especially if you rely on website visits for donations or sign-ups. But for many nonprofits and NGOs, being the trusted source behind the answer still helps your mission. It raises awareness, builds your authority and puts your evidence in front of the people who make decisions.

The simplest place to start is your website analytics. Visits from ChatGPT, Perplexity, Gemini and Copilot often show up as referral traffic in Google Analytics. The numbers might be small, but these visitors tend to be quite engaged.

It’s also worth asking the AI tools yourself every so often. Ask the questions your supporters, journalists or policy makers would ask, and see whether you come up, how your work is described and whether your figures are quoted accurately. There are paid tools that track this for you too, but AI answers change from one conversation to the next, so look at trends rather than individual results.

Where should you start?

Ultimately, optimising for AI search comes down to getting the basics right: clear content, a website that’s technically sound and expertise that’s easy to see. For most nonprofits and NGOs, the expertise is already there. It’s about making sure it’s visible.

If you’re not sure where to begin, I’d start here:

• Check that AI bots can access your website, and decide which ones you’re happy to let in.
• Turn your most important reports into web pages with clear key findings.
• Make sure your key pages answer the most common questions early and in plain language.
• Make your credibility obvious with author bios, methodology and impact figures.
• Ask AI tools about your cause every so often and see what they say.

This is all moving very quickly. What I’ve written here reflects where things stand today, based on the research, platform guidance and our own testing, and I fully expect some of it to change. We’ll keep sharing what we learn along the way.

At Liquid Light, we’ve spent over 25 years helping non-profits, NGOs and purpose-driven organisations build websites that make their work easier to find, understand and trust. AI search is the latest chapter of that story. If you’d like to know how your organisation currently shows up in AI search, or what you could do to improve it, we’d be happy to have a chat.

 

 

• 

  Siddhant Patni

  Account Manager