ChatGPT cannot read haileybury.com
How Haileybury appears when parents and agents build their shortlist through AI assistants, and the three things standing in the way.
Your funnel now has a stage that sits before your website
Families researching UK boarding no longer start on Google and click through ten schools. They ask an assistant to build the shortlist for them, then visit the three or four sites they were handed. Agents in Hong Kong, Lagos, Almaty and Moscow do the same thing, faster.
That means a new gatekeeper sits above your open morning bookings, your enquiry form and your prospectus downloads. If Haileybury is not in the answer, the rest of your admissions funnel never gets the chance to work. Everything below is about whether the machines assembling those shortlists can actually see you.
The short version: you have a strong brand, genuinely good authority and a serious content estate. Three specific, fixable things are stopping that translating into AI visibility, and one of them is not your fault at all.
OpenAI's crawlers are being turned away at your server
Every OpenAI agent receives a 403 Access Denied on haileybury.com
We requested three pages of your site as each major AI crawler. All three of OpenAI's user agents were refused. Every other crawler was let straight in.
This is not in your robots.txt, which permits everyone. The refusal comes from a Varnish caching rule on your hosting layer, so it is almost certainly a security setting that was never intended as a marketing decision.
| Crawler | What it feeds | Homepage | Fees page | How to Apply |
|---|---|---|---|---|
| GPTBot | ChatGPT's knowledge of your school | 403 Blocked | 403 | 403 |
| OAI-SearchBot | ChatGPT Search results and citations | 403 Blocked | 403 | 403 |
| ChatGPT-User | Live fetch when a parent asks about you | 403 Blocked | 403 | 403 |
| ClaudeBot | Claude | 200 OK | 200 | 200 |
| PerplexityBot | Perplexity | 200 OK | 200 | 200 |
| Google-Extended | Gemini and AI Overviews | 200 OK | 200 | 200 |
| Googlebot / Bingbot | Classic search | 200 OK | 200 | 200 |
What this costs you in practice. When a parent asks ChatGPT about Haileybury, it cannot open your fees page, your admissions pathway or your exam results. It answers from directories, forum posts, Wikipedia and third party school guides instead. You lose control of your own story at precisely the moment a family is forming an impression, and any answer it does give is built from sources you did not write and cannot update.
For context, we ran the same test on three comparable schools. Haileybury Melbourne, Berkhamsted and Brighton College all return 200 OK to GPTBot. This block is specific to you.
FIX One hosting rule, roughly a day of work
The block is a server response, not a content problem, so it clears the moment your host allows the three OpenAI agents through. After that, GPTBot needs to recrawl before ChatGPT's picture of Haileybury refreshes, which is why the sooner this is unblocked the better.
A school in Melbourne is out-ranking you for your own name
There are at least six schools called Haileybury: Hertford, Melbourne, Darwin, Almaty, Astana and Malta, plus the Tianjin campus. To a search engine that is one crowded name. To an AI assistant with no clear signal about which is which, it is a coin toss.
The Australian school attracts three times your organic search traffic on lower domain authority than yours (DR 41 against your DR 47). They are not better resourced or better known. They have simply told the machines exactly who they are, and you have not.
| Signal | haileybury.com (Hertford) | haileybury.com.au (Melbourne) |
|---|---|---|
| Homepage title | "Haileybury, a co-educational Independent Day and Boarding School" | "Coeducational Private School Melbourne · Haileybury" |
| Location named in title | No Hertfordshire and UK both absent | Yes Melbourne, first two words |
| School / EducationalOrganization schema | Missing | Present |
| Organization schema | Missing | Present |
| PostalAddress and GeoCoordinates | Missing | Present |
| Structured data actually output | WebSite, WebPage, BreadcrumbList only | School, Organization, PostalAddress, GeoCoordinates |
In plain terms: your site carries no machine-readable record stating that Haileybury is a school, in Hertford, in Hertfordshire, in England, founded 1862, offering IB and A Level. Melbourne's site states all of that in code. When an assistant has to decide which Haileybury a parent means, it reaches for the one that identified itself.
This is the single highest-leverage marketing fix on the list, because it is the difference between an international family being handed your admissions page and being handed a school 10,000 miles away.
585 pages producing the search presence of a much smaller school
Your site has 585 pages and 328 news posts, yet ranks for 276 keywords. Berkhamsted, with a weaker link profile and a smaller estate, pulls 51% more organic traffic than you do. The content exists. It is just not built in a way that search engines or AI assistants can use.
| School | Domain authority | Keywords | Monthly organic visits |
|---|---|---|---|
| Brighton College | 53 | 166 | 6,135 |
| Berkhamsted | 43 | 231 | 4,344 |
| Haileybury | 47 | 276 | 2,879 |
| Oundle School | 50 | 111 | 2,793 |
| Felsted | 42 | 120 | 1,451 |
| St Edmund's College | 32 | 69 | 1,107 |
| Bishop's Stortford College | 42 | 57 | 1,016 |
| Queenswood | 35 | 58 | 595 |
Source: Ahrefs, 8 September 2026. Domain authority is Ahrefs Domain Rating, a 0 to 100 measure of link authority.
Four specific things we found
01 Your fees page description is a raw table dump
Google and AI assistants often quote a page's meta description verbatim. Yours currently reads as unformatted spreadsheet output, which is what a parent sees in a result listing at the exact moment they are price-checking you against Berkhamsted.
02 The How to Apply page describes pastoral care
Your most commercially important admissions page carries a description about pastoral care and community rather than about applying. An assistant asked "how do I apply to Haileybury" reads that description and concludes the page is about something else.
03 The homepage declares six competing H1 headings
A page should make one primary claim. Yours makes six, led by "(I)GCSE Exam Results". Language models weight headings heavily when deciding what a page is about, so the page that should say "leading co-educational boarding school in Hertfordshire" instead reads as a results announcement.
04 Three blog articles, untouched since January
Your thought-leadership library holds three pieces, last modified 5 January 2026. Meanwhile your news feed runs 328 posts. AI assistants cite explanatory content, not announcements. Nobody is asked "what did Haileybury announce", they are asked "IB or A Levels", "is flexi boarding worth it", "how do UK boarding schools support international pupils". Those answers should be yours.
Also missing: an llms.txt file, the emerging convention for telling AI assistants which pages matter and how to describe your institution. It takes an afternoon and it is one of the few places you get to write your own summary.
The maths of a single additional pupil
Senior boarding at Haileybury is £19,820 per term, £59,460 a year including VAT. Senior day is £43,035 a year. A pupil joining in Year 9 and staying to the end of Sixth Form is five years of fee income.
Set against that, none of the work above is expensive. Unblocking OpenAI is a hosting change. Publishing correct schema is a developer day. Rewriting the metadata on your twelve most commercially important admissions pages is a week. The content engine is the part that compounds, and it is the part your competitors have not started either.
This is not a case for a bigger marketing budget. It is a case for pointing part of the existing one at the channel that now sits in front of everything else.
First 90 days
Unblock and identify
Clear the OpenAI 403 with your host. Publish School, Organization, PostalAddress and GeoCoordinates schema. Add llms.txt. Force recrawl. Baseline where you appear across ChatGPT, Gemini, Perplexity and AI Overviews so we can prove movement.
Fix the money pages
Rewrite titles, descriptions and heading structure across admissions, fees, scholarships, bursaries, international and sixth form. Add FAQ schema to the questions parents actually ask. Separate the Hertford entity from the Haileybury Education Group so the name stops competing with itself.
Become the source
Build the answer content assistants cite: IB against A Level, flexi boarding, guardianship, the international admissions journey. Strengthen third party signals in the school guides and directories AI models lean on. Report monthly on citations won, not just rankings.
Everything is measured in the language you already report in: enquiries, open morning bookings, prospectus requests and applications, split by source and by market.
Saigon Digital
We are a digital agency working across web development, SEO and AI visibility, with a long track record in international education. We build and run search programmes for schools and groups across four continents, and we do the same work on our own domain, which is how we found your 403 in the first place.
"Saigon Digital has been our trusted digital partner for over four years. They've delivered a high-performing e-commerce platform that scales globally, supported every market we've launched in, and consistently brought ideas that drive real commercial outcomes. They feel like an extension of our team."
The 403 is the place to start
It is free to fix, it is blocking the largest AI assistant in the world from reading a word you have written, and it has almost certainly been in place for months. We are happy to walk your web team through it whether or not anything else follows.
Talk to Nick Rowe