For two decades, SEO worked because Google indexed the open web and rewarded pages that signalled relevance and authority. Marketers learned the playbook: target keywords, structure pages, build links, optimise for Core Web Vitals. The output was a position in a ranked list of ten links.

That product is being replaced.

Across our client base, between 18% and 34% of high-intent commercial queries now resolve inside ChatGPT, Perplexity, Google's AI Overviews, or Gemini before the user ever clicks a result. The number is climbing every quarter. For brands competing in the Gulf, that shift carries some specific implications that don't translate cleanly from Western SEO advice.

What GEO actually optimises

Generative Engine Optimisation — GEO — is the practice of getting your brand, product, or claim to surface inside an LLM-generated answer. The mechanism is not a ranking algorithm in the SEO sense. It's a retrieval-and-synthesis pipeline:

  1. The model receives a user query.
  2. It retrieves a set of source documents — sometimes from a live web search, sometimes from its training data, often both.
  3. It synthesises an answer from those sources.
  4. It cites — sometimes — the sources it leaned on.

You're optimising for two things: being in the retrieved set, and being weighted heavily enough in the synthesis that your claim makes it into the final paragraph. Neither is the same as ranking on Google.

The five things that move the needle

Across the GEO work we've done with hospitality, retail, and B2B clients in the UAE and KSA, five practices consistently shift visibility:

1. Authoritative, structured claims

LLMs prefer sources that make specific, structured, citable claims. "We are the leading hotel in Dubai" is unhelpful. "Hilton operates four UAE properties with a combined 1,820 keys, the largest of which is Conrad Dubai with 555 rooms" is the kind of sentence that gets retrieved and quoted. Numbers, named entities, and verifiable facts outperform marketing language.

2. Schema, but the right schema

Structured data still matters, but the schema types that LLMs lean on are not the ones legacy SEO tools push. FAQPage, HowTo, Organization, Product with full attribute coverage, and Review with author and rating fields all increase the probability of clean retrieval. Half-implemented JSON-LD with missing required fields gets ignored.

3. Surface area on the right corpora

LLMs are not just reading your website. They lean heavily on Wikipedia, well-known industry directories, news outlets, Reddit, Stack Overflow (for technical queries), and a small set of authoritative review platforms. If your brand is missing from those corpora, you're invisible regardless of how good your own site is. For Gulf brands this is particularly acute — many do not have a Wikipedia entry, are absent from Western directories, and have minimal English-language press.

4. Conversational query coverage

Users don't type GEO queries the way they type Google queries. "Best hotel for a family of four in Dubai with a beach for under AED 1,200" is a single GEO query. The page you want retrieved must be written to plausibly answer that question — which means longer, conversational, comparison-rich content with explicit price ranges, suitability tags, and trade-offs called out.

5. Freshness, but only where it matters

For evergreen queries, LLMs do not strongly prefer recently published content. For time-sensitive queries — pricing, availability, regulations, hours — they do. The freshness signal is binary in a way it isn't for Google. Either your page is dated within a relevant window for the query type, or it gets filtered out of the synthesis.

The Arabic language problem (and opportunity)

Arabic is dramatically under-represented in LLM training corpora. The major models have improved over the last 18 months, but English-language documents still dominate retrieval even for Arabic queries — particularly for niche commercial topics. This creates a specific dynamic in the Gulf:

  • Arabic queries often resolve to English-language sources with the answer translated by the model. The user reads Arabic but the underlying authority is English.
  • Brands that publish high-quality Arabic content with structured data have disproportionate visibility because the corpus is so thin. The competition is much weaker than in English.
  • Models occasionally hallucinate Arabic translations of brand and product names. Brands that publish their own canonical Arabic spellings — and reinforce them through schema — are quoted accurately. Brands that don't get garbled.

The opportunity is asymmetric. Investing in proper bilingual GEO right now produces visibility wins in Arabic queries that are difficult to achieve in English at the same cost.

What is no longer worth doing

Some legacy SEO practices have flipped from useful to negative-ROI in a GEO world:

  • Thin "keyword landing pages." Pages built to rank for specific keyword variations that don't add real informational value. LLMs filter these aggressively.
  • Mass link-building. Generic backlinks from low-authority directories were noise even in late-stage SEO. They are essentially invisible to LLM retrieval.
  • Excessive internal linking. The classic "300 internal links per page" pattern doesn't help GEO retrieval and tends to flatten content quality.
  • Aggressive cookie banners and interstitials. If a crawler can't read your page cleanly, it can't retrieve it. Many Gulf sites have over-eager consent flows that are now actively harming retrievability.

Measuring GEO is harder, but not impossible

The biggest objection we hear is measurement. GEO doesn't show up in Google Search Console. There's no equivalent of organic-keyword reporting. This is true, but solvable:

  1. Synthetic query monitoring. Run a structured set of high-value queries against the major LLMs on a schedule. Track citations, brand mentions, and accuracy of claims. VIMD01 — our data agent — runs roughly 800 such queries weekly across our active client portfolio.
  2. Referrer attribution. Traffic from chatgpt.com, perplexity.ai, and AI Overview clicks is now visible in analytics. The volume is small but growing fast and the intent quality is unusually high.
  3. Brand-mention sentiment. When the LLM does cite you, what is it saying? Are the claims accurate? Is the framing on-brand? This is qualitative work and can't be fully automated yet, but it's the most strategically important signal.

How to start

For a Gulf brand starting from a zero-GEO position, the first 90 days look like this: audit current LLM visibility across 50 priority queries in English and Arabic, fix the structured data gaps, restructure five to ten cornerstone pages for retrievability, claim or extend presence on the corpora that matter (Wikipedia where eligible, industry directories, review platforms), and stand up the synthetic query monitoring loop so you can see what's changing.

That work is roughly four to six weeks of senior strategist time plus continuous agent execution. The results compound: every cycle the model retrieves a cleaner answer, the brand gets more accurately represented, and the cost of further wins drops.

SEO didn't disappear when paid search arrived. It won't disappear when GEO arrives. But the share of attention it captures has already shifted. For brands competing in the Gulf, the window to claim Arabic-language GEO visibility before it gets crowded is narrow. We'd estimate two to three years before the asymmetry closes.