GEO Capability Workshop
for public-sector communications and content teams
A half-day, hands-on workshop that teaches communications and content teams how AI answer engines choose their sources, and how to make official pages the source those engines cite. Every worked example comes from the participating organisation's own pages and questions.
Search ranks your pages. AI chooses what to quote.
When someone asks ChatGPT, Gemini or Google's AI Overviews about a public service or a policy, the engine writes the answer from sources it trusts. Often those sources are not the official pages. The engine quotes a news article, an aggregator or a forum that restated the information, and the restatement is what the reader sees.
For a public publisher this is an accuracy problem, not a marketing one. Restatements go stale, lose nuance or are simply wrong. This workshop teaches teams how AI engines decide what to cite, how to diagnose where their own pages stand, and how to write and structure public information so that the official source is the one quoted.
Nine modules in three movements
Participants diagnose first, then build the signal, then apply and embed it. Every module produces something the team keeps.
Diagnose
The four structural shifts AI search introduces. Why a page can rank first and still never be read. What "citation as currency" means for a public publisher.
Map any question into one of five states: not mentioned, mentioned but not recommended, cited with details wrong, a different source recommended, or recommended cleanly. Then choose the fix that matches the state, not a checklist.
The three pillars: structure, entity authority and answer-first content. llms.txt, AI-crawler access rules, page rendering and firewall behaviour. FAQPage and GovernmentService structured data.
Build the signal
Why AI engines corroborate before they recommend. The earned signals that matter for government: Wikipedia, Wikidata, news media and references from other public agencies. Sequencing earned signals before owned.
Treating GEO as SEO with more keywords. Testing with a single prompt. Confusing AI Overview impressions with citations. Publishing a PDF where an extractable web answer is needed.
Answer first, evidence second, source third. The question patterns that win citations. Writing plainly without losing legal or policy precision.
Apply and embed
Question-shaped headings, an FAQ block, structured data and internal links, applied to a real page. Peer review against citation criteria, not against taste.
Build and lock a set of questions. Track citations across five AI engines each month and sanity-check the tools. Read Search Console AI Overview data without over-reading it.
Ownership across communications, content, web and vendor. A cadence for question review, monthly tracking and escalation. What to do when the fix sits with a third party.
By the end of the workshop, participants will be able to
- Explain how AI answer engines select and cite sources (ChatGPT, Google AI Overviews and AI Mode, Gemini, Perplexity, Copilot) and how this differs from search ranking.
- Diagnose how AI currently treats a question or page using the five GEO States.
- Choose the fix that fits the diagnosis: earned visibility, corroboration, owned content and structured data, or entity positioning.
- Specify the technical requirements for structured data, AI-crawler access and rendering to request from a web team or vendor.
- Prioritise earned signals such as Wikipedia, Wikidata, news media and cross-agency references for a content area.
- Write and rewrite public information in an answer-first format that AI engines can extract, without losing policy precision.
- Apply the Rewrite Framework to turn a live page into a publish-ready brief.
- Run a monthly citation-tracking routine and assign clear owners for ongoing GEO work.
Who should be in the room
Up to 25 participants. The session works best when the people who write the page and the people who publish it are both present.
Communications and content officers
The core audience. They write and own the pages, and every hands-on module is pitched at them. No technical prerequisite.
Digital and web leads
Especially for Modules 3 and 7, which cover the structured-data, crawler-access and template decisions that content officers request rather than make.
Policy officers who review copy
The answer-first rewrite changes where the qualification sits in a sentence. Having the reviewer in the room saves rounds of revision later.
Whoever will own the measurement
Module 8 sets up the tracking sheet live. If that person is absent, the tracking rarely survives the first month.
One organisation per cohort. Running each cohort for a single organisation means the worked examples are that organisation's own pages and questions, not generic case studies.
Before, during and after the session
A half-day only works if the background reading happens before it and the habits are locked in after it.
Sent a week ahead
- A two-page primer: from SEO to GEO, and why it matters for public information
- Confirm access to ChatGPT and Google AI Overviews on a work device
- Each participant submits one question they have asked an AI engine about their own area, and the answer it gave. These are used live in the session
- Pages for the worked examples are agreed with the organisation in advance
Nine modules, one break
- On-site in Singapore or online, recorded for internal reuse
- Roughly two-thirds hands-on or discussion rather than lecture
- Each participant rewrites a real paragraph from their own website
- Live citation tracking on the organisation's own questions
- Ownership (RACI) drafted before close
Lock the habit before it fades
- Each participant commits to one experiment in week one
- The tracking template set up in the session
- A 60-minute group clinic call at day 30
- A second citation check on the same questions, with a before-and-after summary
What every participant keeps
All materials are documents and workbooks the organisation owns outright. Nothing sits behind our tools, and there is no licence to renew.
Participant workbook
The nine modules, the frameworks and the writing patterns, with public-sector examples.
Question set
The set of questions used in the session, as a spreadsheet the team can extend.
Citation scorecard
The monthly tracking template, populated during the session.
Rewrite checklist
A one-page framework participants can apply to any page.
Session recording
For participants who could not attend and for future intakes, with internal re-use rights.
Slide deck
The full deck as a PDF, for internal circulation.
Who delivers it
Terrence Ngu
Founder and Group CEO, Hashmeta. Architect of the 5 GEO States diagnostic and the citation scorecard methodology. More than 15 years in performance marketing across APAC, and he delivers public-sector engagements personally.
Eu Gene
AI and digital marketing educator, master trainer in lifelong learning and NUS Adjunct Associate Professor. Delivers and documents in both English and Chinese.
- 50+ AI SEO and GEO deploymentsin the past six months, across the public sector, education, financial services and e-commerce
- Public-sector benchmark publisherof the 2026 Singapore Public Sector Digital & AI-Visibility Benchmark, covering 71 central-government domains
- Featured in The Edge Singaporeon the new battleground for brand visibility
One fixed fee per cohort
Excluding GST
A fixed fee for one cohort from one organisation. There is no per-participant charge.
- Pre-work: primer, question intake and preparation of the worked-example pages
- The 3.5-hour live session, on-site in Singapore or online
- All participant materials: workbook, question set, scorecard, rewrite checklist and slides
- Session recording with internal re-use rights
- 30-day post-work: the day-30 group clinic and the before-and-after citation check
- All travel and delivery within Singapore
GST is charged at the prevailing rate.
Enquiries
For questions about the workshop content or format, contact Hashmeta.