Insights / Original research
Original research · 2026Who do AI assistants recommend for Egypt travel?
We ran a set of real traveller questions through ChatGPT, Perplexity and Google's AI Overviews and recorded every brand named in the answers. The pattern is consistent, and it isn't good news for operators.
Headline findings
Finding 01
The platforms answer questions about your tours
When a traveller asks an assistant who to book a Nile cruise with, the model returns a shortlist. In our sample, that shortlist was dominated by aggregators and review platforms rather than by the Egyptian companies that actually operate the trips.
This is not a ranking problem. In a generated answer there is no second page and no scrolling past the first three results. A brand is either named or it is absent, and absence is invisible — you never see the booking you didn't get.
Share of citations by source type
Placeholder — replace with measured dataFinding 02
Most operator sites can't be read by the crawlers
For every operator site in the sample we checked three things: whether the main content exists in the raw HTML, whether robots.txt permits the AI crawlers, and whether any structured data identifies the company as a travel operator.
The failure mode was consistent. Newer builds — often the best-looking ones — render their tour content only after JavaScript executes, which means the crawlers receive an effectively empty page. The prettier the rebuild, the more likely it was invisible.
Technical readiness of Egyptian travel sites tested
Placeholder — replace with measured dataFinding 03
What the cited pages had in common
The Egyptian operator pages that did earn citations shared a small set of characteristics, and none of them were about design or brand size.
They answered the traveller's question directly in the opening lines rather than building to it. They carried structured data identifying the company and what it sells. And crucially, they were mentioned somewhere other than their own website — a directory, a review platform, a forum thread, a press mention. Self-assertion alone did not appear to be enough.
Answer position mattered more than page length. Several cited pages were substantially shorter than the operator pages that were ignored.
Characteristics of cited vs. uncited operator pages
Placeholder — replace with measured dataMethodology
How this study was run.
Published in full so anyone can reproduce it, disagree with it, or run it on a different market.
Procedure
- Built a query set of real traveller questions about booking Egypt — cruises, day tours, multi-day itineraries, transfers — in English and Arabic, phrased the way people ask an assistant rather than the way they type into a search box.
- Ran every query through ChatGPT, Perplexity and Google AI Overviews within the same fieldwork window, from a clean session with no personalisation or prior context.
- Recorded every brand and domain named or linked in each answer, classified by source type: OTA, review platform, operator, media, government.
- For every Egyptian operator domain that appeared — and a control set of operators that did not — checked raw HTML content, robots.txt directives, structured data and third-party mention count.
- Compared the cited group against the uncited control on each characteristic.
Limitations, stated plainly
- Assistant outputs are non-deterministic. The same query can return different brands on different runs, so results describe a tendency, not a fixed ranking.
- Answers vary by user location and account history. A clean session is not a neutral session.
- Correlation is not causation. Cited pages shared characteristics; that does not prove those characteristics caused the citation.
- The fieldwork window is a snapshot. Models update, and a repeat run months later would likely differ.
If you're an operator
What to do with this.
The findings point at four actions, in the order that makes each one cheaper than doing it later.
Citing this study
Free to reference and quote with attribution and a link. If you're a journalist or researcher and want the underlying query set or the raw citation log, contact us and we'll send it.
Go Digit'n (2026). "Who do AI assistants recommend for Egypt travel?"
Go Digit'n Research. https://godigitn.com/ai-citations-egypt-travel-study Questions
About the research.
Why does this matter more in travel than elsewhere?
Because travel decisions are open-ended. "Who should I book a Nile cruise with" is exactly the kind of question people now hand to an assistant instead of a search engine — and the answer names two or three companies rather than returning ten links.
Can I reproduce this myself?
Yes, and you should. Take ten queries your customers would actually ask, run them through the three assistants from a clean session, and write down who gets named. It takes an afternoon and it will tell you more than most agency reports.
Do you run this for clients?
Yes — it's the baseline measurement in our GEO programme, repeated monthly so there's a before and after rather than an assertion.
Will you repeat the study?
That's the intention. A single snapshot shows a state; repeated runs show a direction, and the direction is the useful part.
Your own numbers
Find out if the assistants name you.
The free audit includes a citation check on your own money queries across ChatGPT, Perplexity and AI Overviews — so you know where you stand rather than guessing from someone else's study.