01 · What is GEO
The search moment has moved upstream
Generative Engine Optimization is the practice of making an institution visible, accurate, and well-cited in the answers produced by AI search engines: ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude, Copilot.
This is not SEO
SEO optimizes for a ranked list of blue links. GEO optimizes for inclusion in a synthesized answer — typically one or two paragraphs that name a small set of institutions. The competitive question is no longer "where do I rank?" It is "am I one of the universities the AI named at all?"
The scale is real and accelerating
ChatGPT crossed 700 million weekly active users in Q1 2026. Perplexity crossed 35 million MAU. A growing share of prospective students now ask an AI engine to shortlist programs before they ever reach a ranking site or a university's own page — increasingly in their own language.
When an AI engine answers a discovery question — "where can I study a bachelor's in AI in Europe in English?" — it composes an answer from three layers: the training corpus, real-time web search retrieving a small set of authoritative sources, and structured entity graphs. An institution can be invisible in any one of these layers. Algebra Bernays, as the audit shows, is suppressed across all three.
02 · Real-Time Citation Probe
We asked AI where to study AI in Europe. Algebra did not appear. Then we found out why.
We asked "Where can I study a bachelor's in AI and Data Science in Europe, taught in English?" through gpt-5.1 with live web search — the current ChatGPT-generation model. It named a clear shortlist. Algebra was not on it.
So we probed each authority source AI engines trust, by name. The result explains the absence — and it is not what we expected.
| Authority source | What the model actually returned for "Algebra Bernays University" | Recognized |
|---|---|---|
| QS World University Rankings | "not listed… However, Algebra University College is listed." | no |
| Times Higher Education | "not listed… Algebra University College in Croatia is listed." | no |
| Croatian Ministry (MZO) | "not listed… two separate institutions — Algebra University College and Edward Bernays University College — are mentioned." | no |
| Wikipedia | "no dedicated article… no such institution." | no |
| StudyInCroatia (regulator) | "listed… in the Institutions in the Spotlight section." | yes |
03 · Headline Score
8 out of 100 — and the authority AI does see is filed under the wrong name
Orion's education GEO pipeline scored Algebra Bernays and five flagship programs against the same 27-dimension rubric. Every program returned the same near-floor result, for the same reasons: a firewall that blocks AI crawlers, and a rebrand that AI engines never registered.
| Program | Level | URL audited | Score |
|---|---|---|---|
| Data Science & AI | BSc | algebra.hr/.../data-science-and-artificial-intelligence/ | ≈ 8 / 100 |
| Software Engineering | BSc | algebra.hr/.../software-engineering/ | ≈ 8 / 100 |
| Cybersecurity | MSc | algebra.hr/.../cybersecurity/ | ≈ 8 / 100 |
| Digital Marketing | MSc | algebra.hr/.../digital-marketing/ | ≈ 8 / 100 |
| Computer & Data Science | Joint PhD | algebra.hr/.../joint-doctoral-study-programme.../ | ≈ 8 / 100 |
Five identical scores is itself a finding. Every program page shares one template and sits behind one firewall configuration. This is one problem to solve, not five — a single program of work lifts the entire catalog at once, and any new program added later inherits the fix.
Benchmark — how the score breaks down by layer
All scores produced by Orion's education pipeline against the live site. The firewall block zeroed the content layer and triggered a 0.5× penalty on the composite — pre-penalty 16, final 8.
04 · Audit Findings
Top 12 issues identified across the institution
Ranked from lowest to highest impact. Items 10–1 are available in the full engagement briefing. Issue #12 alone is actionable today — it is one setting in a dashboard.
robots.txt welcomes every major AI crawler. The Cloudflare configuration blocks them anyway. When we tested with the real user agents AI engines use, 0 of 5 got through: GPTBot, ClaudeBot, OAI-SearchBot, PerplexityBot, Google-Extended were all blocked at the network layer. This fires the cloudflare_waf ×0.5 penalty that halves the composite, and it is why the content-readiness layer scored zero — the auditor, like the AI engines, never reached a single page. This is almost certainly unintentional: Cloudflare's "AI Scrapers" rule was switched on without anyone realizing it overrides the site's own robots.txt.
WebPage / BreadcrumbList markup. There is no EducationalOccupationalProgram, no Offer (tuition), no educationalCredentialAwarded, no cohort dates, and no FAQPage. The readability layer scored 7 of 100. The content is good; the machine-readable signal is absent.
The full audit is ready.
Issues 9 through 1 — including the critical-severity findings — are covered in detail in the engagement kickoff. Each comes with priority sequencing, effort estimates, and owner assignments.
Schedule the kickoff →05 · Targets
What winning looks like in 12 months
Conservative against the headroom the audit identifies. Aggressive against the current baseline of 13.
| Metric | Today | 6-month | 12-month |
|---|---|---|---|
| Overall AI-search visibility | 8 | 45–55 | 65–70 — peer / top-tier band |
| AI recognizes "Algebra Bernays University" as the entity | No — predecessor names only | Name linked to legacy authority | Recognized + cited under current name |
| Readability (structured data) | 7 | 60 | 80+ |
| Reputation AI engines see (current name) | ~26 | 45 | 75 — Wikipedia + AI-research presence |
| AI crawlers reaching the content | 0 of 5 | 5 of 5 | 5 of 5, sustained |
| Share of AI answers naming Algebra on discovery questions | ≈ 0% | +20 pp | +35 pp — regional shortlist |
| Presence on "AI / data-science program" questions | Absent | Emerging | Named in the regional shortlist |
06 · How Orion Helps
Measurement first. Everything else follows.
Orion is a Generative Engine Optimization platform purpose-built to measure and improve AI-search visibility at the institution level. The audit in this document was produced end-to-end through Orion's education pipeline against the live site.
Entity reconnection
Link the current name to its legacy authority: sameAs + alternateName schema to the predecessor names, a Wikipedia entity with redirects, and update requests to QS, THE, EduRank and the MZO registry.
Reachability + readability
Cloudflare AI-crawler allowlist, llms.txt and sitemap (removes the 0.5× penalty), then one EducationalOrganization block and a program-page schema template across the catalog.
Reputation building
Wikipedia entity seeding, a quarterly press sequence on the partnership constellation, and an AI-research-network presence plan.
Measurement & iteration
Weekly re-audit across the same programs with score deltas and the next-highest-leverage fix surfaced automatically. Citation share tracked per question cluster.
The engagement
| Scope | Institution + flagship program catalog (5 programs audited, extensible) |
| Diagnostic | This audit in full, data exports, and a working session with your operator and academic-communications leads |
| Entity reconnection (lead) | sameAs + alternateName to predecessor names · Wikipedia entity + redirects · QS / THE / EduRank / MZO update requests |
| Reachability fix | Cloudflare allowlist · llms.txt · sitemap |
| Readability build | EducationalOrganization graph · program-page schema template |
| Reputation track | Wikipedia entity · quarterly press · AI-research presence |
| Continuous re-audit | Weekly, with citation tracking on discovery questions |
| Monthly platform fee | €1,500 / month |
| Commitment | 12 months |
| Year-one total | €18,000 |
| Annual prepay option | €16,200 — 10% discount |
Let's talk
The window is open.
Pick a time that works for you.
30 minutes with Simón. We'll walk through the findings shown above, confirm the scope, and tell you exactly what the first two weeks of the engagement look like — starting with the one-setting firewall fix.