Rhino Africa
GEO
Generative engine optimisation
- Role
- UX Architect
- Subject
- Rhino Africa
- Year
- June 2026
- Methods
- IA, GEO, benchmarking
The challenge
How do you structure a page so an AI agent can extract a company's authority and cite them as the expert source?
What it turned out to be
Seven content layers, ordered by the sequence a crawler reads them.
Generative Engine Optimisation is usually treated as a content problem. It is closer to information architecture that happens to get filed under content marketing.
Rhino Africa is Africa's most awarded safari company. Over 5,300 Trustpilot reviews at a 5.0 rating, destination coverage across twelve African countries, and a consultant relationship customers name personally in almost every review.
None of that is declared in a way a machine can read. So I mapped the platform, looked at how AI agents extract and attribute expertise, and worked out what the page hierarchy would have to look like.
Why now
Ranking gets you on the list. Structure gets you in the answer.
Search is shifting from scanning results to reading answers. Perplexity, ChatGPT Search and Google AI Overviews write a response and cite two or three sources.
Page one used to give you a chance at the click. Now the question is whether you are in the answer at all, and that depends on whether the page can be read by a machine.
Method
Understand the ecosystem before touching a page
I wanted to know where a page sits in the system, and what depends on what, before making any page-level decisions.
- 01
Platform information architecture
Map the system. How information is organised, how it flows, and where the seams are.
- 02
GEO performance analysis
Taxonomy depth, locale declarations, multilingual continuity, spatial orientation.
- 03
Persona research
Conversation mining across six review platforms and the client feedback page.
- 04
HNW consumer behaviour
Behavioural psychology and conversion patterns specific to the luxury segment.
- 05
Competitive benchmarking
Six competitors, five UX dimensions, consistently scored.
The system
Six pillars, one destination
Everything funnels toward contact with a Travel Expert. The blog runs on a separate subdomain that mirrors the main taxonomy.
Three structural facts
- 01
The blog is a separate system
It runs on a WordPress subdomain, mirrors the main taxonomy, and cross-links back. Critical top-of-funnel value, and a content consistency risk.
- 02
Tours and Destinations are entangled
Tours are multi-destination packages, and destination pages surface relevant tours. Any IA work has to start with this relationship.
- 03
Trust content is conversion infrastructure
About Us, Price Guarantee, Client Feedback, and Financial Protection sit close to the booking funnel by design. They are not reference material.
Findings
Where it breaks
A lot of this works. The destination taxonomy runs deep, down to reserve level, so Sabi Sand sits under Kruger Private Game Reserve rather than under South Africa. People can arrive at country, region or park level.
5
blog languages
2
locale alternates
On the main site
1
enquiry form language
0
map-based entry points
The language drops away at the moment the user commits
The blog is in five languages. The main site declares two locale alternates. The enquiry form is English only.
What a crawler actually sees
- 01
Authority is never declared machine-readably
No entity block above the fold. No named consultant byline. No schema tying expertise to a person. A crawler reads a hero image and generic intro copy.
- 02
FAQ answers cannot be extracted
They exist, but open with preamble and never name Rhino Africa as the source. An AI agent has nothing clean to lift.
- 03
Geography is the product, but there is no map
Destination filtering is filter-driven by country, duration, and budget rather than spatial. For a platform selling place, that is a structural gap.
Audience
Five persona buckets
Drawn from conversation mining across Trustpilot, Travelstride, TripAdvisor, SafariBookings, Fodor's forums, and the client feedback page.
The Bucket-Lister
PrimaryAffluent, 50+, often couples
Hand-holding and education. They do not know where to start.
The Milestone Celebrator
Honeymoon and anniversary travellers
To feel special and seen. Formulaic responses break trust.
The Experienced Repeat Traveller
Has visited Africa four or more times
Depth and credibility. They come to go deeper.
The Family Organiser
Parent planning a first family trip
Confidence and simplicity. Decision paralysis is high.
The Solo Adventurer
Solo traveller, a growing segment
Autonomy with support. Guidance without being managed.
The consultant is the product.
The finding that reorders every other decision
Across all five buckets, people name their consultant. The platform's job is to get someone to that first conversation without losing them on the way.
The architecture
Seven layers, in crawler reading order
Each layer was tested against a single question. Can an AI agent extract Rhino Africa's authority from this, unambiguously, without human interpretation?
Page Identity (H1)
First signal a crawler reads. Entity and claim in one sentence
[Destination] Safari Guide: Expert Advice from Rhino Africa's Consultants
Authority Block
TravelAgency · aggregateRating · PersonSignals expertise before content begins. The machine reads this first
Entity declaration · Trust signals · Named consultant byline
Structured FAQ (H2 → Q&A)
FAQPageHighest extraction signal. Answers lead with the entity, no preamble
Q: When is the best time to visit? → A: Rhino Africa recommends…
Expert Tips Module (H2)
authorAI looks for named, specific insight. Generic tips are ignored
Attributed consultant recommendations, specific and citable
Lodge Blocks (H2 → H3)
Repeating structured entity blocks, ideal for machine extraction
Name · Location hierarchy · Best for · Season · Consultant note
Conversion Anchor
Human CTA, positioned only after content depth is established
Speak to a consultant about [Destination]. One CTA, no form friction
Machine Layer
FAQPage · Article · TravelAgency · BreadcrumbListInvisible to the user. Tells AI agents who is speaking and why to trust them
Structured data wrapping every layer above
Applied
What changes in practice
Applied to the Kruger National Park destination page. The changes are structural, not cosmetic.
Q: When is the best time to visit Kruger?
“The Kruger National Park is a year-round destination, and there is truly no bad time to experience its magic. That said, many travellers find that the seasons each offer something unique, and much depends on what you hope to see…”
→ Preamble before the answer
→ No entity named as the source
→ Nothing clean to extract or cite
Q: When is the best time to visit Kruger?
“Rhino Africa recommends May–September for Big 5 density; November–April for newborns and dramatic skies.”
→ Entity leads the first sentence
→ Specific, self-contained, citable
→ Wrapped in FAQPage schema
Page identity
Current
“Kruger National Park, A Transformative Luxury Safari”
Rhino Africa is not named. No authority claim.
Recommended
“Kruger National Park Safaris: Expert-Planned Luxury Big 5 Experiences by Rhino Africa”
Entity and authority in the first sentence a crawler reads.
Framework
The nine signals
Each signal is something a machine can detect, and something you can structure a page around.
Structure for extraction
Q&A format, definition blocks, self-contained paragraphs, not intro-heavy prose
Entity clarity
Name brand, location, and subject in full. Never assume context
Authoritative citations
Link to trusted sources; earn citations from credible third parties
Conversational query matching
Write as people ask. 'best time to visit Botswana', not 'optimal visit period'
Schema markup
FAQ, HowTo, Article, LocalBusiness. Improves extraction accuracy
E-E-A-T signals
Named authors with credentials, first-hand accounts, transparent policies
Above-the-fold answer
First 100 words carry the most weight. Answer before context
Brand mentions across the web
Entity recognition reinforced by presence beyond owned channels
Fresh content cadence
Recency is a trust signal. AI favours recently touched pages
E-E-A-T began as E-A-T in 2014. Google added Experience in 2022 to reward genuine, lived involvement over aggregated or AI-generated content.
Competitive set
Looking for the unoccupied ground
Six competitors, three local and three international, scored across five UX dimensions. I was looking for gaps rather than a ranking. Singita, Abercrombie & Kent, and Extraordinary Journeys scored highest at 4.6 average.
Gatsby + Contentful
Next.js SSR
Prismic headless
WordPress
React SPA
WordPress
Scored 1–5 · 1 = weak · 5 = best in class
Four opportunities nobody has taken
Unoccupied across the entire set
- 01
Aspiration-first planning
Only one competitor tries experience-led discovery, and it is closer to how these travellers start planning.
- 02
Transparent enquiry process
Five of six competitors say nothing about what happens after you click Enquire. Three clear steps would cover it.
- 03
Modern technical foundation
Three of six run WordPress. A headless build would be faster and leave room for personalisation.
- 04
Itinerary visualisation
Nobody offers an itinerary builder or a shareable journey map. Once someone is talking to a consultant, a plan they can both see and change is the obvious gap.
Deliverable
Six implications, ordered by leverage
Where targeted architecture work has the greatest effect on both experience and commercial conversion.
Surface the consultant early
Across all five persona buckets, people name their consultant. That is the product, not a support function.
Distribute trust signals
Awards, press, named consultants, and testimonials belong at every decision point, not quarantined on dedicated pages.
Remove friction from enquiry
One clear step rather than a form. This audience does not wait around.
Extend multilingual continuity
Language has to carry from the blog through to the enquiry form.
Add a spatial orientation layer
The product is geography and there is no map. That is a structural gap.
Build personalisation cues
Signal that the trip is being planned for this person, not assembled from a template.
Outcome
The output is a structure rather than a screen. It makes fifteen years of expertise readable to a machine that has never met a consultant.
Reflection
What I took from it
- 01
GEO gets filed under content marketing. Most of it is information architecture
- 02
Machines cannot infer authority. If expertise is not declared explicitly and consistently, it does not exist to a crawler
- 03
Mapping the ecosystem first meant the page decisions were mostly already made
- 04
The first hundred words carry the most weight. Answering before the preamble helps readers and crawlers equally