PRIME is the Government Communications Service’s new five-stage framework for keeping public information accurate, discoverable and trusted in AI search: Plan, Respond, Implement, Monitor, Evaluate. It was published on 20 October 2026 and is written for government communicators, but every public sector organisation whose residents ask AI tools about services, benefits or safety can use it.

Ask an AI tool when a VAT payment is due, and it gives you an answer. Not a list of links. An answer. If that answer is wrong, the person asking has no reason to doubt it.

That’s the problem the Government Communications Service (GCS) set out to tackle in its new guidance, Strengthening public trust in the age of AI-powered search, published on 20 October 2026. It introduces the PRIME framework to help public bodies keep their information visible, accurate and trusted wherever people now search.

One finding from the guidance is worth pausing on. When the Office for National Statistics (ONS) assessed how AI Overviews handled its data, it was cited in 97% of the responses analysed and named in 62% of them. But ONS subject matter experts judged only 21% of those responses to be fully accurate. Being found is not the same as being right, and for a public body, being wrong has real consequences.

This matters well beyond central government. Open and honest communication is the foundation of public trust. If people receive inaccurate information about your organisation before they ever reach you, everything else you do to build that trust becomes harder.

In short: AI-mediated search means tools that generate direct answers instead of listing links, and GCS is worried that official information is being missed, outdated or misrepresented when they do.

The guidance defines AI-mediated search as search engines and tools that use artificial intelligence to give direct answers. That includes AI Overviews and AI Mode in Google, Microsoft Copilot, ChatGPT, Claude, Perplexity and others. The practice of making content easy for these tools to find, understand and represent accurately is called generative engine optimisation, or GEO.

The numbers behind the shift, as set out in the guidance:

• Search is a daily habit. An Ofcom report on online habits found 65% of UK adults look up information online every day.
• Clicks are disappearing. Research from Studio 36 Digital, cited by GCS, suggests over 54% of UK searches now end without a click to any website.
• AI search use is growing fast. In the YouGov Great Britain AI Search Report 2026, also cited in the guidance, 71% of AI search users said they were using it more than a year earlier, and just 5% said less.
• Younger audiences look elsewhere too. In the same YouGov report, 49% of Gen Z use social media to search, and 37% use online communities or forums.
• Queries are changing. Google’s own research shows the average query in AI Mode is around three times longer than a traditional search, and more than one in six AI Mode searches now involve images, audio or video.

GCS is clear that this is a starting point, not a rulebook. It describes GEO as a fast-moving discipline and says the guidance will evolve with it.

In short: PRIME is five stages, applied together with your digital, policy and content teams, to keep content accurate and AI-ready on an ongoing basis.

In short: find out what your audiences are actually asking, then audit and retire anything outdated, duplicate or wrong.

GCS recommends starting with search insight. Tools like Google Trends can show the language people really use. The UK Health Security Agency (UKHSA) did exactly this, mirroring the phrasing people searched for and testing whether updating existing posts or publishing new ones moved the needle. Their case study reports an increase in blog views of 160%.

Then comes the content audit. The guidance is explicit that this goes well beyond your website. It should cover press notices, factsheets, campaign sites and speeches, and also social media posts, influencer content and third-party sites. Prioritise high-traffic and high-priority content first rather than waiting for a perfect full audit.

Two practical notes. GCS says teams cannot yet archive their own GOV.UK content, and that the Government Digital Service is developing a solution, so keep archiving outdated content on your non-GOV.UK sites and channels in the meantime. And if you use AI tools to help with the audit, always have someone who knows your communications review the output.

In short: write so that both people and machines get the answer straight away.

The guidance leans on GOV.UK writing standards. Use clear, descriptive headings that say what the section covers and match what people are searching for. Answer the question in a single line directly beneath the heading. Put supporting detail below that, and background last. Write in plain English, keep sentences short (the GOV.UK style guide suggests under 25 words), and use bullet points where they help. You can read the full standard in GOV.UK’s guidance on clear structure, and for new content, GOV.UK’s guidance on writing for user needs is a useful companion.

That advice is as useful for a pinned social post, a profile bio or a reply template as it is for a web page. If residents keep asking the same thing, the clearest answer should exist in one consistent form wherever they might find it.

In short: make sure the content you keep can be crawled, indexed and understood by machines.

This stage is about the technical side: content that is accessible to everyone, indexable by search engines, supported by sensible site structure, and marked up with schema.org structured data so automated tools can read it accurately. GCS also points teams to Google’s own AI optimisation guidance. If your SEO foundations are strong, you’re already helping your GEO, and the guidance recommends Google’s SEO starter guide and the PRCA’s guide on how to make your content fit for AI search as further reading.

Be realistic about ownership here. This is largely a job for your website and digital teams, and the guidance stresses working with them early. The most useful thing a comms team can do is make sure social profiles and posts stay consistent with the pages they point to, so the same facts appear everywhere.

In short: use the GCS Evaluation Cycle’s Inputs, Outputs and Outcomes stages, and treat AI search optimisation as always-on.

The guidance asks teams to frame evaluation around stages one, two and four of the GCS Evaluation Cycle: Inputs, Outputs and Outcomes. In practice:

• Inputs: your processes, such as archiving outdated pages, assigning page owners, using structured data, and tracking how long updates take to appear in AI outputs.
• Outputs: how AI platforms cite and summarise you, including citation frequency, answer inclusion rate, share of voice against third-party sources, and referral traffic from AI platforms. The ONS case study shows how one organisation approached this.
• Outcomes: whether AI tools represent your information accurately, keep key facts consistent when paraphrased, and genuinely answer people’s questions.

There’s no mature measurement standard yet, so GCS recommends combining automated reports with manual spot-checks across several AI engines, and not relying on a single tool. It also suggests giving direct feedback, such as using the thumbs-down on Google AI Overviews, when you spot an inaccurate answer.

If you want to go deeper on the evaluation side, see how the GCS Evaluation Cycle maps onto the OASIS framework.

PRIME doesn’t replace anything. It sits alongside the frameworks GCS and the wider public sector already use:

• OASIS is for planning campaigns. PRIME’s Plan and Evaluate stages echo its audience insight and scoring steps.
• RESIST is for responding to misinformation. PRIME’s Monitor stage is the same early-warning instinct, applied to AI search.
• SAFER is for checking an environment is appropriate before you publish. PRIME asks a related question about where your content lives and how it’s represented.
• The GCS Evaluation Cycle is the measurement backbone PRIME explicitly builds on.

Taken together, they all serve the same end: public sector organisations that communicate accurately, listen well and can show their impact are the ones that earn lasting trust.

PRIME is mostly about your content and your web estate. Orlo’s role is the social and community part of it:

• Social Listening and Media Monitoring show how people talk about your organisation on social media and in the news, so you can spot and correct inaccurate claims there.
• A unified inbox surfaces the questions people ask most often, so you can answer them clearly and consistently.
• Content calendar and approvals help teams keep wording aligned across departments and channels.
• Social Archiving keeps a record of your social activity.
• Trust Indicator and Performance Analytics help you evaluate outcomes and outputs on the social side.

What Orlo does not currently do is monitor Reddit or AI search tools, track how often AI platforms cite you, or manage the structured data on your website. For those, pair Orlo with your digital team and the kind of dedicated tools listed in the GCS guidance, and build regular manual checks into your routine.

Over 400 public sector organisations use Orlo. If you’d like to see how it could support your approach to PRIME, book a demo.

• PRIME (Plan, Respond, Implement, Monitor, Evaluate) is GCS’s new framework for keeping public information accurate, discoverable and trusted in AI search, published on 20 October 2026.
• Being found by AI is not the same as being represented correctly. ONS was cited in 97% of responses analysed, but only 21% were judged fully accurate.
• The guidance covers social media posts, influencer content and third-party sites, not just your website.
• HMRC’s “HMRC Verified” social alert shows how a fast, clear social response can correct inaccurate AI search advice.
• Evaluate continuously using Inputs, Outputs and Outcomes from the GCS Evaluation Cycle, and don’t rely on a single tool.
• PRIME builds on frameworks like OASIS, RESIST and SAFER rather than replacing them.