GEO, AEO and AI Search Glossary for Arab Markets: 70 Terms Explained

This glossary defines 70 GEO, AEO and AI search terms for Arab markets. GEO means getting cited and named in AI answers, AEO means structuring pages so engines can lift them as answers, and Google treats both as SEO. Every entry pairs a quotable definition with a note for Saudi, UAE and Egyptian teams, checked against sources from October 2026.
Generic English glossaries skip the problems that decide Arabic visibility: hamza spelling, dialect, right-to-left markup, and analytics that lose WhatsApp and ChatGPT clicks. We built this one from VOCTOS’s own October 2026 tests of Arab websites and from Google, OpenAI, Anthropic and W3C documentation.
How to use this glossary. Terms sit in six groups, each opened by a short overview. Read the first paragraph under a term on its own: it is the definition, and it is the text we mark up for search engines. The line marked “In Arab markets” adds the local angle and links to our deeper study where one exists.
Jump to a group: AI search terms, crawling and retrieval, Arabic language and script, measurement, local search and reputation, ads in AI answers, or the 10 most confused terms.
AI search and answer engine terms
AI search terms describe how assistants and Google’s AI features write answers and which brands those answers name. An agency proposal can list GEO, AEO and AIO as three services, yet Google calls optimising for its AI features plain SEO. The 17 entries below also separate a citation from a mention, which AI visibility tools count as different metrics.
Generative engine optimization (GEO)
Generative engine optimization (GEO) is the work of getting a brand cited and named in answers written by AI systems such as ChatGPT, Gemini and Google’s AI Overviews. It extends SEO rather than replacing it.
In Arab markets: Arabic is used by 0.6% of websites with a known content language (W3Techs, read 7 October 2026). Engines answering Arabic questions therefore draw on a thin pool of Arabic sources. Our guide to generative engine optimization sets out the method.
Answer engine optimization (AEO)
Answer engine optimization (AEO) is the practice of structuring content so a search or AI system can lift it as a direct answer. It began with featured snippets and voice assistants and now covers AI answers too.
In Arab markets: Google’s guide to generative AI search says optimising for its AI features is still SEO. Treat AEO as a writing discipline: answer first, one question per section, in the Arabic your buyers type. Our AEO explainer shows the page patterns.
AI optimization (AIO)
AI optimization (AIO) is a loose umbrella label for any work that improves how AI systems find, read and present a brand. The same three letters also abbreviate Google’s AI Overviews, so ask which one a proposal means.
In Arab markets: When an agency report shows an “AIO” number, check its source. An AI Overview count comes from Google Search alone, while a broader AI score blends prompts run in ChatGPT, Gemini and Perplexity.
| Term | Goal | Where results show | How it is measured |
|---|---|---|---|
| SEO | Rank pages in search results | Google and Bing results pages, including Google’s AI features | Rankings, impressions and clicks in Search Console |
| AEO | Get a passage lifted as the answer | Featured snippets, voice answers, AI answers | Snippet ownership, AI impressions |
| GEO | Get cited and named in AI-written answers | ChatGPT, Gemini, Perplexity, AI Overviews, AI Mode | Citations, mentions and share of voice across a fixed prompt set |
| AIO | Either of the above, or AI Overviews specifically | Depends on who uses the term | Ask for the definition before reading the number |
Large language model (LLM)
A large language model (LLM) is an AI model trained on very large amounts of text to predict and generate language. The models behind ChatGPT, Gemini and Claude are LLMs, and on their own they answer from training data rather than from today’s web.
In Arab markets: Regional models count too. Saudi Arabia’s HUMAIN Chat, built on the ALLaM 34B model, launched on 25 August 2025, and Qatar has Fanar. Test brand answers there as well as in ChatGPT and Gemini.
AI Overviews
AI Overviews are Google’s AI-written summaries that appear at or near the top of some search results, with links to the pages that support them. Google shows them only when it judges them helpful beyond the classic results, so many queries never trigger one.
In Arab markets: Google brought AI Overviews to Arabic in 2025, across the region at once. Ahrefs measured a 58% lower click-through rate for the top result when an AI Overview appeared (300,000 keywords, desktop, December 2025). Our AI Overviews optimisation guide covers what gets cited.
AI Mode
AI Mode is Google’s conversational search tab, where the whole results page is an AI answer with supporting links and room for follow-up questions. Google positions it for comparisons and questions that need reasoning, and lists query fan-out among its techniques.
In Arab markets: AI Mode has worked in Modern Standard Arabic since 8 October 2025. Search Console counts its impressions together with AI Overviews in one number, so the two cannot be split in the report. Our AI Mode and SEO guide explains the difference.
ChatGPT search
ChatGPT search is the ChatGPT feature that retrieves current web pages and shows them as source links inside the answer. OpenAI builds its index for it with OAI-SearchBot, a different crawler from GPTBot, which gathers training data.
In Arab markets: ChatGPT sent between 72% and 83% of AI chatbot referrals in Saudi Arabia, the UAE, Egypt, Qatar, Kuwait, Oman and Bahrain in September 2026 (StatCounter). Gemini rose fastest in those seven markets, from between 1.2% and 2.3% in August 2025 to between 13.3% and 21.4% in September 2026.
AI agent
An AI agent is an AI system that carries out a task for a person, such as comparing products or booking an appointment, by browsing sites and taking actions. Google notes that browser agents read screenshots, page structure and the accessibility tree.
In Arab markets: An agent booking a clinic slot in Riyadh or ordering in Cairo needs labelled form fields and readable buttons in Arabic and English. A booking flow that exists only inside WhatsApp gives it nothing to complete.
Query fan-out
Query fan-out is the technique in which an AI search system splits one question into several related searches, runs them at once and merges the results into one answer. Google lists it among the techniques behind AI Overviews and AI Mode.
In Arab markets: Map the sub-questions behind your Arabic head terms, such as price, location, eligibility and reviews, and answer each one on the page. Google warns that mass-producing pages for every fan-out variant breaks its scaled content abuse policy.
Citation (AI citation)
An AI citation is a link or named source that an AI answer attaches to a claim, showing which page it drew on. A citation can send a click, and a page can also shape an answer without being cited.
In Arab markets: Semrush’s AI Visibility Toolkit reports mentions, cited pages and citations as three separate metrics. Check which one a vendor quotes. Our research on Arabic AI citations has the regional detail.
Brand mention
A brand mention is an AI answer naming your brand in its text, with or without a link to your site. Mentions shape what buyers remember, while citations show which pages the engine used.
In Arab markets: Run each buyer question in English, Modern Standard Arabic and your customers’ dialect. Log the brands each answer names. Google says chasing inauthentic mentions does not help in its AI features.
Share of voice in AI answers
Share of voice in AI answers is your brand’s portion of all brand mentions across a fixed set of tracked prompts, compared with named competitors. It is only as meaningful as the prompt set behind it.
In Arab markets: A prompt set written only in English leaves out the questions Arab buyers type in Arabic. Build it in English, Modern Standard Arabic and Gulf or Egyptian dialect. Then keep it fixed month to month.
AI visibility score
An AI visibility score is a vendor’s single number summarising how often a brand appears in AI answers across a prompt set. Semrush, for example, scores it from 0 to 100 against competitors, and each tool calculates it differently.
In Arab markets: Ask the vendor how many Arabic prompts, and in which dialects, sit behind the number before you compare it month to month. Our AI visibility score method covers the prompt testing behind it.
Prompt tracking
Prompt tracking is running a fixed list of buyer questions in AI assistants at regular intervals and logging which brands and pages each answer names. It is the AI-search counterpart of rank tracking.
In Arab markets: Answers move fast. When we re-ran seven ChatGPT shopping prompts a week apart, on 4 and 11 October 2026, the top pick changed on three of them. Date every run and keep the account and location fixed.
Zero-click search
A zero-click search is a search that ends on the results page without a click to any website, because an answer, map or AI summary already satisfied the searcher. AI answers raise the share of searches that end this way.
In Arab markets: Search Console’s AI report counts impressions but no clicks, so a page can gain AI visibility while its clicks stay flat. That gap is the zero-click effect. Compare AI impressions with organic clicks for the same pages and dates.
Hallucination
A hallucination is a false statement that an AI model presents as fact, such as a wrong price, opening hour or product feature. It comes from how models generate text, and grounding in live sources reduces it without removing it.
In Arab markets: In our 11 October 2026 ChatGPT test, the same air fryer was quoted at SAR 439 in an English answer and at 329 to 415 riyal in a Gulf Arabic one. Keep prices and branch hours identical on your Arabic and English pages.
Featured snippet
A featured snippet is a block in Google’s classic results that quotes a passage from one web page to answer a question, with a link to that page. It predates AI Overviews and quotes a single source instead of writing a summary.
In Arab markets: Arabic pages win snippets with the habit that also helps in AI answers. Put a direct answer of 40 to 60 words under a heading phrased as the question, in the words people type.
How do AI engines find, read and cite a page?
AI engines find pages through crawlers, read them through retrieval systems that pick passages at answer time, and identify brands through entity databases. Each layer has its own control: robots.txt and firewall rules, Search Console settings, and consistent naming. On 2 October 2026, 6 of 37 Arab brand robots.txt files we read blocked at least one AI crawler.
Retrieval-augmented generation (RAG)
Retrieval-augmented generation (RAG) is a method in which an AI system first retrieves relevant documents, such as pages from a search index, and then writes its answer from them. Google says its AI features in Search work this way.
In Arab markets: Retrieval runs on the search index. An Arabic page that Google has not indexed, or cannot show with a snippet, cannot appear as a supporting link in AI Overviews or AI Mode. Fix indexing on Arabic URLs first.
Grounding
Grounding is the step that ties an AI answer to sources fetched at the moment of the question, so the model writes from current pages instead of memory alone. Google’s documentation treats grounding and RAG as the same technique.
In Arab markets: Two separate switches govern grounding on Google: the Search generative AI control in Search Console for AI Overviews and AI Mode, and Google-Extended for Gemini Apps and Vertex AI.
Entity
An entity is a uniquely identifiable thing, such as a company, person, place or product, that search and AI systems treat as one item across its different names. Entities let engines connect facts about your brand across pages.
In Arab markets: An Arab brand often has three names in use: Arabic script, an English name and a Latin transliteration. Use the same set everywhere and list them in your Organization markup; on 3 October 2026, 5 of 15 Arabic homepages with JSON-LD named the brand only in Latin script.
Knowledge Graph
The Knowledge Graph is Google’s database of entities and the relationships between them, which feeds knowledge panels and helps Google work out who or what a query is about. The Knowledge Graph is the database, and a knowledge panel is one display of it.
In Arab markets: Search your brand in Arabic script and in English. If only one search shows a panel, treat it as a sign that Google has not linked both names to the same entity yet.
Structured data (schema markup)
Structured data is code, usually JSON-LD using the schema.org vocabulary, that labels facts on a page such as an organisation’s name, address, products or prices. Google says its AI features need no special markup, though structured data still earns rich results.
In Arab markets: On 3 October 2026, 13 of 28 Arabic homepages we checked carried no JSON-LD at all. All 4 banks had none. Prices in markup must use the digits 0 to 9 even when the page displays Arabic-Indic digits.
AI crawler
An AI crawler is a bot that fetches web pages for an AI company. Vendors split crawlers by job: training models, indexing pages for AI search, and opening a page when a user asks, each with its own token.
In Arab markets: On 2 October 2026, 25 of 37 readable robots.txt files from large Arab brands named no AI crawler, and 8 of 37 homepages showed a bot challenge or blank page to an ordinary browser. Check the firewall as well as the file, as our robots.txt study of 50 Arab brands explains.
| Token | Vendor | Job | What blocking it does |
|---|---|---|---|
| GPTBot | OpenAI | Collects training data | Keeps pages out of future OpenAI training; ChatGPT search is unaffected |
| OAI-SearchBot | OpenAI | Indexes pages for ChatGPT search | Removes the site from ChatGPT search answers, apart from navigational links |
| ChatGPT-User | OpenAI | Opens a page a user asked about | OpenAI does not commit to robots.txt for these visits; use a firewall rule |
| ClaudeBot | Anthropic | Collects training data | Keeps pages out of Claude training |
| Claude-SearchBot | Anthropic | Improves Claude’s search results | Reduces the site’s visibility in Claude’s search answers |
| Claude-User | Anthropic | Opens a page a user asked about | Stops Claude retrieving the page for user questions |
| PerplexityBot | Perplexity | Indexes pages for Perplexity answers | Keeps the site out of Perplexity’s search index |
| Perplexity-User | Perplexity | Opens a page a user asked about | A robots.txt block is usually ignored, per Perplexity; use a firewall rule |
| Google-Extended | Control token for Gemini training and grounding | No effect on Google Search, AI Overviews or AI Mode | |
| Googlebot | Crawls for Google Search, including its AI features | Removes the site from Google Search entirely |
GPTBot
GPTBot is OpenAI’s crawler for collecting pages that can be used to train its future models. Blocking GPTBot keeps your pages out of OpenAI’s training data without removing you from ChatGPT search, which relies on OAI-SearchBot.
In Arab markets: In our 2 October 2026 check of large Arab brands, 8 of 37 readable robots.txt files named GPTBot and 2 blocked it. Blocking it is a training decision.
OAI-SearchBot
OAI-SearchBot is OpenAI’s crawler that indexes pages so ChatGPT search can show and cite them. OpenAI says sites that block it are left out of ChatGPT search answers, apart from navigational links.
In Arab markets: None of the 37 readable Arab brand robots.txt files we read on 2 October 2026 blocked OAI-SearchBot, and 6 named it. Keep it that way if you want ChatGPT to cite your pages.
ChatGPT-User
ChatGPT-User is the user agent OpenAI uses when ChatGPT opens a page because a person asked it to, rather than to crawl the web. OpenAI does not commit to following robots.txt rules for these user-initiated visits.
In Arab markets: Count ChatGPT-User hits in your server logs next to OAI-SearchBot hits. A rise in both, alongside more Direct visits to deep pages, is consistent with ChatGPT sending people your way.
ClaudeBot
ClaudeBot is Anthropic’s crawler for collecting training data for Claude models. Anthropic runs two more agents, Claude-SearchBot for search quality and Claude-User for pages a person asks Claude to read, and says all three honour robots.txt.
In Arab markets: The older tokens anthropic-ai and Claude-Web still appeared in 6 and 4 of 37 Arab brand robots.txt files on 2 October 2026. Anthropic’s current crawler page lists neither, so those rules do not cover ClaudeBot.
PerplexityBot
PerplexityBot is Perplexity’s crawler for indexing pages that its answer engine can cite, and Perplexity says it does not collect content for training foundation models. Its separate Perplexity-User agent fetches pages when a person asks and usually ignores robots.txt.
In Arab markets: Perplexity sent 4.25% of AI chatbot referrals in Saudi Arabia in September 2026, against 5.85% worldwide (StatCounter), yet 2 of 37 readable Arab brand files block PerplexityBot.
Google-Extended
Google-Extended is a robots.txt control token, with no crawler of its own, for Google’s Gemini models. It decides whether your content can train future Gemini models and ground answers in Gemini Apps and Vertex AI, and it has no effect on Google Search.
In Arab markets: Blocking Google-Extended leaves AI Overviews and AI Mode untouched, because those follow Googlebot rules and the Search generative AI control. Google’s crawler documentation says the token is not a ranking signal, and 2 of 37 readable Arab brand files block it.
Blocking Google-Extended leaves AI Overviews and AI Mode untouched, because those follow Googlebot rules and the Search generative AI control.
llms.txt
The llms.txt file is a proposed Markdown document at a site’s root that lists a site’s main pages for AI tools, suggested by Jeremy Howard in September 2024. Google Search ignores it, and OpenAI, Anthropic and Perplexity do not say their crawlers read it.
In Arab markets: Only 4 of 37 readable Arab brand sites served an llms.txt on 2 October 2026, and only 1 of the 3 files we opened contained any Arabic text. Our llms.txt study of Arab websites shows what they contain.
Search generative AI control
The Search generative AI control is a Search Console setting that includes or excludes a site’s links and content from Google’s AI Overviews, AI Mode and generative features in Discover. Google rolled it out to all sites on 31 August 2026.
In Arab markets: Inclusion is the default. A child property can override its parent, so check every Search Console property a past agency or developer created. Google’s help page for the control says exclusion takes one to two days and does not stop AI training.
Arabic language and script terms
Arabic adds problems that English search does not have: letters with several written forms, digits in two scripts, right-to-left layout and dialects that change the words in a query. These 16 entries decide whether Arabic pages match what people type, and none of the English or German glossaries we reviewed covers them.
Modern Standard Arabic (MSA)
Modern Standard Arabic (MSA), fusha in Arabic, is the shared written standard of news, government and formal content across Arab countries. It is the form Google named when it launched AI Mode in Arabic on 8 October 2025.
In Arab markets: MSA pages read correctly in every Arab market, so they suit brands that sell in Saudi Arabia, the UAE and Egypt from one Arabic site. Keep dialect for headings, FAQs and ads where buyers type colloquial words.
Dialect targeting
Dialect targeting is choosing which spoken variety of Arabic, such as Gulf, Egyptian, Levantine or Maghrebi, a page or ad should match, based on the words the audience actually types. It is usually layered on top of an MSA base.
In Arab markets: Read search-term reports by country before writing a dialect version. Keep facts, prices and claims identical. Start with short commercial queries and AI prompts, where people write the way they speak.
Gulf Arabic queries (Khaleeji)
Gulf Arabic queries are searches typed in the dialects of Saudi Arabia, the UAE, Kuwait, Qatar, Bahrain and Oman, with vocabulary that differs from MSA and from Egyptian Arabic. A Saudi shopper writes wesh for “what”, where MSA uses matha.
In Arab markets: When we asked ChatGPT six shopping questions in Gulf Arabic on 11 October 2026, two answers came back partly or wholly in Egyptian Arabic. Check the dialect of AI answers, not only the brands they name.
Egyptian Arabic queries (Masri)
Egyptian Arabic queries are searches written in Egyptian colloquial Arabic, whose vocabulary and spelling differ from both MSA and Gulf dialects. Egyptians write ayez for “I want” and dilwa’ti for “now”, where Saudis write abgha and al-heen.
In Arab markets: Egyptian searchers see AI answers at least as often as Gulf ones. In our September 2026 Search Console data, 25.8% of Egyptian impressions on a large Arabic health site carried an AI link, against 23.8% from Saudi Arabia.
Arabizi (Franco-Arabic)
Arabizi is Arabic written in Latin letters and digits, with numbers standing in for sounds English lacks: 3 for ain, 7 for a strong h and 2 for the hamza. Egyptians often call it Franco.
In Arab markets: Gulf reviewers write in MSA, dialect, English or Arabizi, so review replies and brand monitoring have to read all four. If buyers spell your brand in Arabizi, add that spelling to keyword research and to alternateName in your markup.
Code-switching
Code-switching is mixing Arabic and English in one query or sentence, such as an Arabic question wrapped around an English product or model name. Search and AI systems have to match both scripts, so pages should carry the names buyers type.
In Arab markets: Keep model names such as iPhone or Galaxy in Latin script on Arabic pages, and add the Arabic-script spelling once if Search Console shows buyers typing it.
Hamza and alef normalisation
Hamza and alef normalisation treats alef written with hamza above, hamza below or madda as plain alef. Searches typed with or without the hamza then match the same words, a rule Arabic retrieval research has used since at least 2002.
In Arab markets: Semrush data for September 2026 shows 22,200 monthly Saudi searches for “car insurance” spelled with the hamza and 12,100 without it. In our 8 October 2026 test, 14 of 20 Arab online stores returned the same results for both spellings; our Arabic site search study lists the fixes.
Taa marbuta
Taa marbuta is the letter that ends many Arabic feminine nouns, written like a haa with two dots. Users often type it as a plain haa, so well-built search systems match words such as sharika and sharikah (company) as one.
In Arab markets: Taa marbuta was the weakest pair in our 8 October 2026 test: only 10 of 20 Arab online stores returned the same results for both spellings. Google Ads lists misspellings as close variants but does not document this pair.
Alef maqsura
Alef maqsura is the dotless final letter that looks like a yaa, as in mustashfa (hospital) and the word ala (on). It is often swapped with a dotted yaa, especially in Egyptian writing, so normalisation treats the two endings as one.
In Arab markets: A site search that misses the swap shows “no results” for a correctly typed product. Test your 20 best-selling product names with both endings before a sale season such as White Friday.
Diacritics (tashkeel)
Diacritics, tashkeel or harakat in Arabic, are the small marks above and below letters that show short vowels and doubled consonants. Everyday Arabic leaves them out, so search systems strip them before matching.
In Arab markets: Add diacritics only where a word is ambiguous without them. Keep them out of URLs, product names and ad keywords, where they create a second spelling of the same word.
| Spelling pair | What changes | Example (transliterated) | Treat as |
|---|---|---|---|
| Alef with hamza and bare alef | Hamza above or below the alef dropped | ta’min and tamin (insurance) | Bare alef |
| Alef with madda and bare alef | Madda dropped | aakhir and akhir (last) | Bare alef |
| Taa marbuta and haa | Final letter typed as haa | sharika and sharikah (company) | One letter (haa) |
| Alef maqsura and yaa | Dotless and dotted final forms swapped | mustashfa and mustashfi (hospital) | One letter (yaa) |
| With and without diacritics | Short-vowel marks added or left out | Vowelled and bare spellings of one word | Strip the marks |
| Arabic-Indic and Western digits | Numeral forms | U+0660 to U+0669 and 0 to 9 | Western digits |
Arabic-Indic digits
Arabic-Indic digits are the numeral forms used in much Arabic text, Unicode U+0660 to U+0669, as opposed to the Western digits 0 to 9. Prices, phone numbers and model numbers appear in either form.
In Arab markets: In our 8 October 2026 test, 3 of the 19 Arab online stores we could check returned zero products when a number was typed in Arabic-Indic digits. Normalise both digit sets in site search, forms and phone fields.
RTL and the dir attribute
RTL (right-to-left) is the writing direction of Arabic, declared in HTML with the dir attribute. Setting dir=”rtl” on the html element lays out text, tables and punctuation from right to left, and the W3C advises markup rather than CSS for it.
In Arab markets: On 2 October 2026, 3 of 27 Arabic homepages we checked left dir off the html element. Phone numbers, prices and English names inside Arabic lines also need bidi isolation, or they display in the wrong order.
Language codes (ar, ar-SA, ar-EG, ar-AE)
Language codes label a page’s language, and optionally its region, in lang and hreflang attributes. Arabic is ar, and ar-SA, ar-EG and ar-AE add Saudi Arabia, Egypt and the UAE; Google accepts only ISO 639-1 language and ISO 3166-1 Alpha 2 region codes.
In Arab markets: Google ignores dialect codes such as arz (Egyptian Arabic), ar-KSA is invalid, and sa on its own means Sanskrit. Use plain ar unless prices, stock or legal text change by country.
hreflang for Arabic pages
The hreflang annotation tells Google which language and regional versions of a page exist, so it can show each searcher the right one. Every version must list itself and all others, and Google ignores pairs that do not link back.
In Arab markets: On 2 October 2026, 9 of 29 Arab brand homepages had no hreflang in the page head and 7 of 20 sets lacked x-default. Google’s localized versions guide also says it ignores hreflang and lang when detecting a page’s language. Our hreflang, lang and dir audit has a working template.
Arabic URL slugs and percent-encoding
An Arabic URL slug is the readable part of a web address written in Arabic words, which browsers and servers transmit percent-encoded. Each Arabic letter becomes six characters, such as %D9%85, so encoded slugs grow long fast.
In Arab markets: Google accepts Arabic slugs, and its URL documentation encodes Arabic words as an example. In our 2 October 2026 sample, encoded Arabic URLs averaged 235 characters against 71 for English slugs; our Arabic URL slug study weighs the trade-offs.
Transliterated brand names
A transliterated brand name is a brand written in another script, such as an Arabic-script spelling of an English name or a Latin spelling of an Arabic one. Each spelling is a separate search query, and often a separate entity until you link them.
In Arab markets: Google Business Profile allows one name and bans repeating it in a second language or script. Bilingual names belong on the website and in Organization markup, not in the profile name field.
Seen your own site in this list? Check my Arabic pages against these 70 terms
Measurement terms for AI and search traffic
AI answers change what your reports can see. Google counts AI impressions but no clicks, ChatGPT links now arrive without a referrer, and bots inflate page views in most Arab markets. The 10 entries below show where each visit lands in GA4 and Search Console, and which numbers to report as floors rather than totals.
AI impressions (Search Console)
AI impressions are the times a link to your site was shown inside an AI Overview or AI Mode answer, as counted in Search Console’s Generative AI performance report. The report breaks them down by page, country, device and date, with no clicks.
In Arab markets: In September 2026, 24.8% and 27.6% of Google web impressions on two Arabic health sites we measured carried an AI link, and 35.5% of the clinic site’s Saudi impressions did. The Generative AI report help page explains the counting, and our Search Console AI report study has the country table.
AI referral traffic
AI referral traffic means the visits that reach a website after someone clicks a link inside an AI assistant such as ChatGPT, Gemini or Perplexity. Clicks from Google’s AI Overviews and AI Mode count as Google organic traffic instead.
In Arab markets: The volume is still small. AI chatbots sent 0.41% of tracked referral visits in Saudi Arabia in September 2026, against 82.45% from search engines (StatCounter).
AI referral share
AI referral share is each assistant’s portion of the website clicks that AI chatbots send. StatCounter measures it from referrals to sites running its tracker, so it shows who sends visits, not how many people use each assistant.
In Arab markets: In September 2026, ChatGPT sent 75.00% of AI chatbot referrals in Saudi Arabia, 80.99% in the UAE and 72.16% in Egypt. Gemini sent 16.54%, 13.25% and 21.39%, and Copilot, Perplexity and Claude shared the remaining 8.46%, 5.74% and 6.43%. Worldwide, the split was 80.48%, 10.91% and 8.59% (StatCounter, read 2 October 2026).
AI assistant channel (GA4)
The AI assistant channel is a default channel group that Google Analytics 4 introduced in 2026 for sessions whose referrer matches Google’s list of AI assistants. GA4 labels those sessions with the medium ai-assistant.
In Arab markets: Google’s channel page names ChatGPT, Gemini, DeepSeek, Copilot and Grok, but not Perplexity or Saudi Arabia’s HUMAIN Chat. Treat the channel as a floor. Our GA4 AI traffic guide adds a custom channel group.
Direct traffic inflation
Direct traffic inflation is growth in the Direct channel caused by visits that arrive with no referrer and no campaign tag. Many clicks from AI answers and messaging apps land there, so Direct no longer means only bookmarks and typed URLs.
In Arab markets: It stacks with WhatsApp, where pasted links also arrive without a referrer, and with phone use: mobile carried 68.27% of Saudi web traffic in September 2026 (StatCounter). Our ChatGPT Direct traffic guide shows how to estimate the hidden visits.
UTM parameters
UTM parameters are tags added to a link’s URL, such as utm_source, utm_medium and utm_campaign, that tell analytics tools where a visit came from. They take priority over the referrer, but only on links you or a platform tag.
In Arab markets: ChatGPT appended utm_source=chatgpt.com to answer links from June 2025, and by our 15 October 2026 check the tag was gone. For ChatGPT Ads you add your own UTMs, and OpenAI appends only its oppref click ID.
Referrer policy
A referrer policy is the rule that controls how much of the previous page’s address a browser passes to the next site. Sites set it with an HTTP header, a meta tag or a link attribute, and it decides whether analytics sees a referral.
In Arab markets: When we checked chatgpt.com on 2 October 2026, it set strict-origin-when-cross-origin, which passes only the origin and not the full URL, yet its answer links added noreferrer on top. The link attribute wins.
Noreferrer (rel attribute)
Noreferrer is a link attribute value, written rel=”noreferrer”, that tells the browser to send no referrer when the link is followed. Without a UTM tag, analytics on the destination site cannot see where the click came from and files it as Direct.
In Arab markets: All 18 ChatGPT answer links we checked on 15 October 2026 carried noreferrer and none had a UTM tag, across English, Gulf Arabic and Egyptian Arabic prompts. Adobe’s Cedric Huesler counted about 31,000 fewer daily ChatGPT visits and 26,000 more direct ones across roughly 500 pages over 9 to 13 October.
Bot traffic
Bot traffic is the share of requests or page views generated by automated software, from search and AI crawlers to scrapers and fake-click bots, rather than people. Left unfiltered, it inflates sessions and drags down engagement and conversion rates.
In Arab markets: StatCounter recorded bots at 30.85% of tracked page views in Saudi Arabia, 45.07% in the UAE and 20.08% in Egypt in September 2026, against 31.13% worldwide. Filter before you report.
Server log analysis
Server log analysis is reading your web server or CDN logs to see which bots and browsers requested which URLs, and when. It is the only first-party record of AI crawler visits, because most crawlers never run analytics tags.
In Arab markets: Verify AI crawler hits by user agent and by the vendor’s published IP ranges, since user agents are easy to fake. Sites behind Cloudflare also need the CDN’s bot logs, because an edge block never reaches the origin server.
Local search and reputation terms
For clinics, restaurants, showrooms and branches, the first answer a buyer sees is often a map result or a star rating, long before the buyer reaches a website. Brand searches add a second layer: an Arab brand has one results page for its Arabic name and another for its English name.
Google Business Profile (GBP)
Google Business Profile is Google’s free listing for a business location, holding the name, address, hours, phone, photos and reviews shown in Maps and the local pack. Google’s AI guide names it as a source of local details in AI responses.
In Arab markets: Google allows one name per profile, the real-world name, and bans repeating it in a second language or script even when the shop sign shows both. Saudi addresses also need the National Address entered in Google’s own field order.
Local pack
The local pack is the block of map results, usually three businesses with a map, that Google shows for searches with local intent such as “dentist near me”. Google ranks it mainly on relevance, distance and prominence.
In Arab markets: Google’s local ranking help page says prominence draws on how many websites link to a business and how many reviews it has. Distance is fixed. Branches in Riyadh, Dubai and Cairo compete on relevance and prominence.
NAP consistency
NAP consistency means a business’s name, address and phone number match exactly across its website, Google Business Profile, directories and social profiles. Conflicting versions make it harder for search and AI systems to treat them as one business.
In Arab markets: Arabic and English versions of an address count as one NAP only when the details agree: the same building number, district and phone format. Write phone numbers in one international format, such as +966, on every page.
Review velocity
Review velocity is the rate at which a business gains new reviews, such as reviews per month. It is an industry metric rather than a Google term, and Google’s local ranking page names the number of reviews under prominence.
In Arab markets: Steady requests after every visit beat bursts. Google’s Maps policy bans incentivised and gated reviews, and a profile caught faking engagement can stop receiving new reviews for a period and show a public warning.
Review gating
Review gating is asking only satisfied customers for a public review, or screening customers first and sending unhappy ones to a private form. Google’s Maps content policy bans it, along with paid and incentivised reviews.
In Arab markets: Saudi Arabia and the UAE handle fake reviews through general false-advertising rules. The Saudi E-Commerce Law caps fines for false electronic claims at SAR 1 million, and UAE rules fine misleading advertising AED 250,000. This is general information, not legal advice.
Brand SERP
A brand SERP is the full results page Google shows when someone searches a brand name, including the site, knowledge panel, reviews, news, social profiles and any AI Overview. It is the first impression for anyone checking you before buying.
In Arab markets: Arab brands have at least two brand SERPs, one for the Arabic-script name and one for the English or transliterated name. Audit both. Since late September 2026, signed-out Maps visitors see the star summary but not individual reviews.
Knowledge panel
A knowledge panel is the information box Google shows for an entity in its Knowledge Graph, with facts such as logo, founders, address and social profiles. Google generates it automatically, and verified owners can suggest changes.
In Arab markets: A business with a Business Profile gets a local panel from that profile, while a brand panel comes from the Knowledge Graph, and the two are managed in different places. Keep the Arabic and English names identical.
Search engine reputation management (SERM)
Search engine reputation management (SERM) is the work of shaping what search results and AI answers say about a brand or person. It strengthens accurate pages and profiles and deals with false or harmful results through policy and legal routes.
In Arab markets: Google says Gemini uses Maps reviews to ground answers, so unanswered Arabic reviews now feed AI answers as well as star ratings. Our search engine reputation management guide covers the process.
Google says Gemini uses Maps reviews to ground answers, so unanswered Arabic reviews now feed AI answers as well as star ratings.
Ads inside AI answers
Paid placements now sit inside or beside AI answers, and each platform decides separately who can buy and who sees the ads. For an Arab advertiser the split matters: ChatGPT Ads are open to Gulf companies through self-service, while Google’s drug-ad test in AI Mode covers the US only.
ChatGPT Ads
ChatGPT Ads are paid placements that OpenAI shows below ChatGPT answers, labelled as sponsored, and sells through its Ads Manager or sales team. OpenAI shows them to Free and Go users and logged-out visitors, not to paid plans.
In Arab markets: Saudi and UAE companies joined OpenAI’s self-service Ads Manager list in mid-September 2026, but OpenAI blocks health, financial and legal services ads outside the US. Our ChatGPT Ads Manager guide for Saudi Arabia and the UAE covers eligibility.
Sponsored answers
A sponsored answer is any paid placement shown inside or next to an AI-generated answer and labelled as an ad, from ChatGPT’s sponsored units to ads in Microsoft Copilot. Labels and placement rules differ by platform.
In Arab markets: Microsoft’s help page says all eligible campaign and ad types are automatically opted in to serving in Copilot. Advertisers cannot opt out, so review any Microsoft Advertising campaign you run in the Gulf with that in mind.
Ads in AI Overviews and AI Mode
Ads in AI Overviews and AI Mode are Google text and Shopping ads shown with Google’s AI-generated answers and labelled as sponsored. Google described both placements as tests when it spoke to Adweek in October 2026.
In Arab markets: The October 2026 expansion to some pharmaceutical ads covers the US only, and no Arab country appears on any of Google’s three drug-ad location lists. We found no Google statement on when these placements reach Arabic results.
Performance Max
Performance Max is a Google Ads campaign type that runs one goal across Search, YouTube, Display, Discover, Gmail and Maps from a single campaign. Google’s automation chooses placements, bids and ad combinations from the assets and audience signals you supply.
In Arab markets: The campaign language setting stops applying to Performance Max’s Search inventory but still governs its YouTube, Display, Discover and Gmail placements, so keep Arabic and English assets in separate asset groups.
AI Max for Search
AI Max for Search is a Google Ads feature set for Search campaigns that adds AI-driven matching beyond your keywords, text customisation from your site and final URL expansion. Advertisers switch it on campaign by campaign.
In Arab markets: Google says AI Max text customisation supports all Google Ads languages, Arabic included, but its AI campaign localisation beta lists 10 European target languages and no Arabic as of October 2026.
Which 10 terms do Arab marketing teams confuse most?
The ten terms Arab teams confuse most are GEO, AIO, citation, AI Overviews, Google-Extended, GPTBot, llms.txt, the Knowledge Graph, AI impressions and Direct traffic. Each confusion leads to a costly decision, from blocking the wrong crawler to reading AI impressions as clicks.
| Term | What it is | What it is not |
|---|---|---|
| GEO | Work to get cited and named in AI answers, built on SEO | A separate channel that replaces SEO |
| AIO | Either “AI optimization” or Google’s AI Overviews, depending on the writer | One agreed metric |
| Citation | A link or named source attached to an AI answer | A brand mention without a link |
| AI Overviews | Google’s AI summary inside the classic results page | AI Mode, Google’s separate conversational tab |
| Google-Extended | A robots.txt token for Gemini training and grounding | A switch for AI Overviews, AI Mode or rankings |
| GPTBot | OpenAI’s training crawler | The crawler behind ChatGPT search, which is OAI-SearchBot |
| llms.txt | An optional file proposed for AI tools | A ranking or citation signal in Google Search |
| Knowledge Graph | Google’s database of entities | The knowledge panel, which displays it |
| AI impressions | Times your link was shown in an AI Overview or AI Mode answer | Clicks or visits |
| Direct traffic | Visits with no referrer or tag, now including many AI and WhatsApp clicks | Only bookmarks and typed URLs |
Our read
Arab teams need the old vocabulary applied to Arabic more than they need a new one. In our October 2026 checks, 16 of 20 Arab online stores failed at least one Arabic spelling pair, 13 of 28 Arabic homepages had no JSON-LD and 9 of 29 had no hreflang in the page head.
We treat GEO and AEO as SEO with better measurement, and in Arab markets that work starts with Arabic handling. Before paying for AI visibility tracking, a site should normalise Arabic spelling in its search, declare lang, dir and hreflang correctly, and let OAI-SearchBot, PerplexityBot and Googlebot through the firewall.
Reporting needs the same care. A ChatGPT click with no referrer lands in Direct, so a falling ChatGPT row in GA4 can hide steady demand. Our position is to report AI visits as a measured floor plus a labelled estimate, and to read Search Console AI impressions next to clicks rather than as a win on their own.
FAQ
Is GEO just SEO with a new name?
Mostly, yes. Google says optimising for its AI features is still SEO, and its AI Overviews and AI Mode need no special markup or files. GEO adds two things worth paying for: tracking how AI assistants mention and cite you, and fixing the Arabic-specific gaps that keep your pages out of their answers.
Do I need an llms.txt file on my Arabic website?
No for Google, optional elsewhere. Google Search ignores llms.txt, and OpenAI, Anthropic and Perplexity do not say their crawlers read one. Only 4 of 37 readable Arab brand sites published the file in our October 2026 check. Robots.txt rules, firewall settings and indexable Arabic pages matter far more for AI citations.
Our ChatGPT traffic in GA4 dropped in October. Did we lose visibility?
Probably not. In our 15 October 2026 check, every ChatGPT answer link carried noreferrer and no UTM tag, so GA4 files those clicks as Direct. Compare new users landing on deep pages through Direct before and after 7 October, when GPT-6 became the default, and count ChatGPT-User hits in your server logs.
Should our Arabic pages be written in MSA or in dialect?
Write the core page in Modern Standard Arabic, because it reads correctly in every Arab market and it is the Arabic Google named for AI Mode. Add dialect where buyers use it: headings, FAQs, ad copy and search-term variants. Keep prices, branch details and claims identical across every version you publish.
Where can I see AI Overview traffic for my site in Search Console?
Open Performance, then the Generative AI report, which Google rolled out to all sites on 31 August 2026. It counts how often links to your pages appeared in AI Overviews and AI Mode, by page, country, device and date. It shows impressions only, with no clicks and no query data.
Where should a team start?
Start with the terms that change a decision this quarter, not all 70. For most Arab teams that means four: Search Console AI impressions, Direct traffic inflation, the AI crawler tokens and Arabic normalisation. Each one changes a report, a robots.txt file or a site search setting within a week.
Knowing what GEO stands for matters less than knowing whether engines can find, match and count your Arabic pages. If Search Console already shows the Generative AI report and your robots.txt names AI crawlers on purpose, move on to prompt tracking in Arabic and English. If either is missing, fix crawler access, Arabic spelling and hreflang first, because no AI visibility tool can count citations for pages engines cannot read.
Want a second pair of eyes on the basics? Book a review of my AI search and Arabic SEO setup
Everything else we have run on GEO & AEO
Written by whoever ran the work, not a content team
38 articles







