Conversational discovery27 min read · Sources & credits
In this chapter

This chapter explains how AI assistants build an answer, where they find their sources, where and how ads are placed beside those answers, and what the click-through research says about the traffic that results. The headline finding is that answers and ads are run as separate systems: OpenAI states that "Ads do not influence the answers ChatGPT gives you" and that advertisers never receive chats, while independent US studies show that AI summaries reduce clicks to websites, with Pew recording 8% of visits clicking a traditional result when an AI summary appeared against 15% without one. No equivalent study exists for the UAE.

In brief

  • ChatGPT search rewrites a question into one or more targeted queries, sometimes uses partner search providers, and states "There is no way to guarantee top placement" (OpenAI Help Center). Google's AI features may use "query fan-out" across subtopics.
  • Questions in Google AI Mode are nearly three times longer than traditional searches, according to Google.
  • ChatGPT ads appear below the response, carry the advertiser's name and favicon, and are selected by a relevance-weighted second-price auction (OpenAI Help). Targeting uses the current conversation and, depending on country and settings, broader ChatGPT context (OpenAI).
  • Advertisers never receive chats, chat history, memories, name, email, precise location or IP address, only aggregated views and clicks (OpenAI Help). No ads appear to users under 18, in Temporary Chats, or in mental and personal health conversations (OpenAI; OpenAI Help).
  • Separate measurements put sponsored placements in about a quarter of eligible ChatGPT responses: 26% of responses to Free and Go users (Similarweb, recorded 17 August 2026, geography not stated) and 25.94% of 50,006 commercial prompts (SE Ranking, US, July 2026) (PPC Land; Search Engine Journal).
  • ChatGPT ad CTR from 30 March to 16 September 2026 was 0.73% in the US and 0.74% in the UK, with no Middle East data published (PPC Land).
  • On Google, Ahrefs found AI Overviews correlate with a 58% lower CTR for the top-ranking page; Seer found organic CTR down 61% and paid CTR down from 19.70% to 6.34% on AI Overview queries, while brands cited in the overview did better than uncited brands.
  • UAE media law requires ads to be "clearly identified, distinct, and separate" from other content, and 65% of UAE respondents say they are willing to rely on information from an AI system (UAE Legislation portal; KPMG Lower Gulf).

Explain the customer journey in ordinary language

A person may begin with a problem, add constraints and compare alternatives over several exchanges. For example, an adult learner could ask about changing careers, then compare certification routes, study schedules and local providers. A business buyer could begin with a workflow problem and only later request software options. These examples illustrate possible journeys; they are not observed ONS conversation data.

Platform data support the general pattern. Google says AI Mode users ask questions nearly three times longer than traditional search queries; its MENA blog put it as "two or three times the length of traditional search queries" when AI Mode launched in Arabic. The two statements are close but not identical. OpenAI says about one in five show commercial intent: Al Bayan relayed it as about 20% of ChatGPT conversations globally, while TNW quoted OpenAI's Colin Fleming as 20% of one billion weekly users. Most conversations are therefore not shopping conversations, and an ad system has to decide when not to show anything.

Conversation depth affects advertising too. Similarweb data read from a chart show worldwide CTR highest at the first turn of a conversation, approximately 0.66%, falling to approximately 0.36% to 0.48% at turns 5 to 21 (30 March to 30 August 2026) (PPC Land). Similarweb reports that about a third of ChatGPT ad impressions occur in the first response of a session.

The useful planning unit is the decision the customer is trying to make. Identify what the buyer needs to know, which claims the business can support and what action would sensibly follow. This produces more useful creative than repeating a list of commercial keywords without context, and it fits a platform that offers no keyword targeting, only context hints (The Decoder, 2026; OpenAI).

How an assistant builds an answer and chooses sources

Assistants answer in two broad ways. They can answer from what the underlying model learned in training, or they can search the web at the moment of the question and summarise what they retrieve. Retrieval is what gives a business website a chance to be cited, and it follows rules each platform publishes.

ChatGPT search

ChatGPT search is available to Free, Plus, Team, Edu and Enterprise users, including logged-out Free users, and searches automatically when a question "might benefit from information on the web" (OpenAI Help Center). It rewrites the user's question into one or more targeted queries, "sometimes partners with other search providers", and ranks results on "a number of factors"; OpenAI states "There is no way to guarantee top placement". "Any public website can appear in ChatGPT search", provided it does not block OAI-SearchBot. A page disallowed by robots.txt may still appear as a link and title where there are "signals the page is relevant to a user's query" (OpenAI Help Center). ChatGPT adds utm_source=chatgpt.com to referral links automatically, which makes organic ChatGPT traffic visible in analytics (OpenAI Help Center).

Google AI Overviews and AI Mode

Google's AI features may use a "query fan-out" technique, running multiple related searches across subtopics and data sources. A supporting link must be "indexed and eligible to be shown in Google Search with a snippet", and "There are no additional technical requirements". Google states "You don't need to create new machine readable files, AI text files, or markup to appear in these features"; the controls are the existing nosnippet, data-nosnippet, max-snippet and noindex, and AI feature clicks are reported in Search Console under the "Web" search type. Google-Extended controls the use of content to train Gemini models and "does not control Search features"; Google said in January 2026 it is "exploring updates" to let sites opt out of Search generative AI features, with no launch date. A Google analyst was reported as saying Google "doesn't support LLMs.txt and isn't planning to" (Search Engine Land).

Perplexity

PerplexityBot is "designed to surface and link websites in search results on Perplexity" and "is not used to crawl content for AI foundation models". Its user-triggered fetcher "generally ignores robots.txt rules" because a user requested the fetch (Perplexity Docs).

Crawler or agentPlatformPurposeControlSource
OAI-SearchBotOpenAISurfaces sites in ChatGPT searchAllow in robots.txt to be included; changes take effect in about 24 hoursOpenAI Help Center; OpenAI [Platform]
GPTBotOpenAITrainingDisallow to exclude content from trainingOpenAI Help Center; OpenAI [Platform]
ChatGPT-UserOpenAIUser-initiated visits; not automatic crawling"robots.txt rules may not apply"OpenAI [Platform]
OAI-AdsBotOpenAIChecks landing pages submitted as ChatGPT ads; not used for trainingMust not be blocked for ads to runOpenAI; OpenAI Help [Platform]
Googlebot (Search)GoogleIndexing for Search, including AI Overviews and AI Modenosnippet, data-nosnippet, max-snippet, noindexGoogle Search Central [Platform]
Google-ExtendedGoogleUse of content to train Gemini modelsDoes not control Search featuresGoogle [Platform]
PerplexityBotPerplexitySurfaces and links sites in Perplexity results; not for foundation-model trainingPerplexity recommends allowing itPerplexity Docs [Platform]
Perplexity-UserPerplexityUser-requested fetches"generally ignores robots.txt rules"; changes up to 24 hoursPerplexity Docs [Platform]

Which sources get cited

Citation data are early and mostly from the US. Similarweb reports that ChatGPT's citation presence on US prompts rose from about 1.6% in June 2025 to about 6.8% in May 2026; Travel and Hospitality prompts carry a citation about 23% of the time and Professional Services under 4% (citation rates, not ad rates) (Similarweb, 2026). Pew found Wikipedia, YouTube and Reddit together made up 15% of sources cited in Google AI summaries and 17% in standard results, in US data from March 2025 (Pew Research Center). BrightEdge reports that 34% of AI citations come from sources brands can influence through PR and nearly 10% from social platforms. None of these figures covers Arabic content or UAE queries.

What it means for marketersThe same robots.txt file can block an ad from running and a site from being cited. A UAE business preparing for ChatGPT Ads should check that OAI-AdsBot and OAI-SearchBot are not blocked, and should treat GPTBot, the training crawler, as a separate decision (OpenAI Help; OpenAI).

Keep the answer system and ad system distinct

OpenAI describes ads as separate from ChatGPT responses. Its first published principle reads: "Ads do not influence the answers ChatGPT gives you" (OpenAI). Its head of monetisation said ads "do not influence the core organic model" (Axios). Advertisers cannot shape or rank those responses through ad spend. The advertising system uses relevance signals that include conversational context and advertiser material; additional personalisation depends on the user's settings. Advertisers do not receive private conversations (OpenAI).

Where the ad appears and what it contains

The ChatGPT ad unit sits below the response and shows the advertiser's name, favicon, title, description, landing page and image (OpenAI Help). Adobe Advertising's ChatGPT ad settings give hard limits of 50 characters for the title and 100 for the description; about 16 and 32 characters is what typically displays. OpenAI's sponsored agents test, run with select US advertisers from 16 September 2026, lets a user who clicks an ad open a labelled conversation with a business-sponsored agent that is separate from ChatGPT's own answers (Reuters via Zawya, Anzar Mehraj). OpenAI has also tested placing several advertisers in one sponsored placement (MarTech, Constantine von Hoffman, 2026).

How the ad is chosen

Eligible ads are chosen through "a relevance-weighted, second-price auction", with bids set at ad-group level (OpenAI Help). Objectives are Views (billed per thousand impressions), Clicks (per click) and Conversions (optimised CPC or CPM); "you don't pay per conversion" (OpenAI Help; OpenAI). OpenAI states that "OpenAI's ads system controls all delivery decisions". Advertisers supply context hints, which are relevance signals rather than delivery controls and which cannot enforce geography or exclusions (OpenAI). There is no keyword targeting; The Information reported in April 2026 that advertisers receive broad topic tags (The Decoder).

What context the ad system uses

OpenAI says targeting uses the current conversation and, depending on country and user settings, broader ChatGPT context. TechCrunch reported at the US launch that ads were matched on conversation topic, past chats and previous ad interactions. With personalisation turned off, only the current thread is used (OpenAI Help). Personalisation rules vary by region: in the US it is on by default with an opt-out, while the EU and UK require opt-in consent, and at the European launch on 24 August 2026 personalisation based on conversation context, general location and device was withheld (Crypto Briefing via Gizmodo; TechWyse, DJ Kennedy). OpenAI advises against custom audiences for the EEA and Switzerland until personalisation is available. No source states which personalisation default applies to UAE users (see Gaps).

Privacy controls and exclusions

OpenAI's privacy commitments for ads are among the most specific published by any platform. "We keep your conversations with ChatGPT private from advertisers, and we never sell your data to advertisers". Advertisers never receive chats, chat history, memories, name, email, precise location or IP address, and receive only aggregated views and clicks (OpenAI Help). Ads data are deleted from servers within up to 30 days (OpenAI Help). Brand-safety vendors DoubleVerify and IAS are piloting suitability controls that assess safeguards without access to conversations (Adweek, Trishla Ostwal, 2026).

Users have choices. OpenAI says it will "always offer a way to not see ads in ChatGPT", and Free plan users can choose an ad-free option with "lower usage limits and reduced feature access". Paid plans from Plus upwards are ad-free (OpenAI). No ads are shown where a user says, or OpenAI predicts, they are under 18; logged-out users see ads "designed to be suitable for all ages"; and Temporary Chats carry no ads.

Some conversations are excluded entirely. OpenAI's ad policy v1.6, updated 10 September 2026, places no ads on conversations in sensitive categories, including regulated goods such as gambling or tobacco, or on "vulnerable user-model interactions": emotionally reliant contexts, "Mental and personal health conversations" and sensitive user journeys. The policy's April 2026 change means medical, legal and financial advice contexts are no longer categorically blocked from ad placement, but that concerns where ads may appear, not which advertisers may buy them (OpenAI). Outside the US, ads for health services are generally prohibited (OpenAI).

How the other platforms separate ads from answers

Google's new AI Mode formats, Conversational Discovery ads and Highlighted Answers, include "an independent AI explainer" and stay labelled "Sponsored". Ads within AI Overviews are text and Shopping ads, drawn from existing campaigns, and are excluded from sensitive categories (Google Ads Help). Business Agent for Leads, announced for "coming months", would place a brand chat agent inside the ad in place of a static lead form (Google). Direct Offers place exclusive offers in AI Mode responses as a pilot (Google; Google, Jan 2026). Perplexity's ads were sponsored follow-up questions (PPC Land, 2024). Amazon's Rufus ads are sponsored prompts billed per click (Shopifreaks, 2026). Meta's approach does not place ads inside Meta AI at all; it uses Meta AI conversations as a targeting signal for ads elsewhere, excluding sensitive topics.

PlatformWhere the ad sitsSeparation from answerSignals usedUser and privacy controlsSource
ChatGPTUnit below the response; sponsored agent after click (US test)"Ads do not influence the answers"; labelled sponsored agent separate from answersCurrent conversation; broader context depending on country and settings; context hintsAd-free option; no ads under 18 or in Temporary Chats; chats never shared; data deleted within up to 30 daysOpenAI Help; OpenAI; Reuters via Zawya, Anzar Mehraj [Platform]
Google AI OverviewsAbove, below or within the overviewWithin-overview ads excluded for sensitive categories; no segmented reportingExisting Search, Shopping and Performance Max signalsStandard Google ad controlsGoogle Ads Help [Platform]
Google AI ModeIn-response formats in test"Independent AI explainer"; labelled "Sponsored"Broad match, AI Max, Performance Max, ShoppingNo AI Mode opt-out for advertisersGoogle; Adthena [Platform] [Industry]
Microsoft CopilotAds in Copilot responses; Showroom ads on purchase intentNot described in sources reviewedPurchase intent expressed in conversationNot described in sources reviewedMicrosoft Advertising [Platform]
PerplexitySponsored follow-up questionsProgramme winding down over trust concernsNot describedNot describedPPC Land; Campaign [Industry]
Meta AINo ads inside Meta AI; signals feed ads on Meta appsNot applicableMeta AI interactions; sensitive topics excludedNo opt-out per Reuters; UK, EU, South Korea excludedMeta Newsroom; Reuters via TradingView; Nasdaq [Platform] [Industry]
Amazon RufusSponsored prompts in the assistantNot describedShopping contextPrompt-level reporting for advertisersModern Retail; Shopifreaks [Industry]

This has a practical consequence: a campaign brief should describe the offer accurately, without claiming access to a user's private questions. ONS should never sell a service as monitoring individual ChatGPT conversations, because no such access exists (OpenAI Help). The landing page must earn the enquiry by answering the customer's next question clearly.

How often ads appear, and how relevant they are

Independent measurements of ChatGPT ad load come from panels and prompt tests, mostly in the US. They agree on the order of magnitude, an ad in roughly one in four or five responses or questions, but use different populations and methods (PPC Land; Search Engine Journal).

StudyFindingPopulation and periodSource
SimilarwebSponsored placements in 26% of responses to Free and Go users; CTR 0.50%Recorded 17 Aug 2026; geography not statedPPC Land [Industry]
Similarweb (AI Search Stats)About 26% of ChatGPT responses contain an ad; about a third of ad impressions in the first response of a sessionPopulation not stated; data to May 2026Similarweb [Industry]
SE Ranking25.94% of 50,006 commercial prompts returned an ad; healthcare niche 28.69% vs 2.64% on Google AI Mode; 14.35% of ad impressions no closer to the prompt than randomUS, collected 23 Jul 2026; AI Mode figure from a separate SE Ranking report (30 Jun), not like for likeSearch Engine Journal [Industry]
WiredAbout 1 in 5 questions in a new thread triggered an ad; travel most frequent; ads sometimes promoted a competitor of a brand named in the questionUS mobile, 500 questions, c. Mar 2026Search Engine Land, Anu Adegbola [Industry]
Similarweb and GraphiteShare of ad-eligible users who saw at least one ad: UK 45.7%, US 43.4%, Brazil 34.9%, Mexico 27.8%Nine markets, Sep 2026PPC Land [Industry]
Sensor TowerAds per user per hour on the US mobile app up 163% in August vs April 2026US mobilePPC Land on Sensor Tower [Industry]

The SE Ranking result on relevance is the most useful for planning: 14.35% of ad impressions were no closer to the prompt than a random ad would be (Search Engine Journal). Together with Wired's observation that an ad could promote a competitor of a brand the user had named, it shows that relevance matching is imperfect and that a business may appear beside, or be displaced by, a competitor in its own brand conversations (Search Engine Land, Anu Adegbola, 2026). OpenAI is testing exclusion targeting, meaning contexts an advertiser wants to avoid, with a small group of advertisers, but the exclusion categories have not been published (eMarketer).

Click-through: what the studies show

Two questions matter for a business: whether AI answers reduce clicks to websites in general, and how often people click the ads placed beside them. All the studies below are from the US or global panels, and all are correlational.

Clicks from AI summaries in Google

Pew Research Center tracked the browsing of 900 US adults in March 2025. Users clicked a traditional search result in 8% of visits where an AI summary appeared, against 15% where none appeared, and clicked a link inside the summary in 1% of visits. 18% of the 68,879 unique searches produced an AI summary, and users ended their browsing session on 26% of visits with a summary against 16% without (Pew Research Center, 2025).

US users clicked a result in 8% of visits with an AI summary, against 15% without

Share of Google search visits with a click, %, US, 900 adults, March 2025

US users clicked a result in 8% of visits with an AI summary, against 15% without

Source: Pew Research Center, 22 Jul 2025

Ahrefs, analysing desktop Search Console data for 300,000 keywords, found that the presence of an AI Overview "correlates with a 58% lower average clickthrough rate" for the top-ranking page, with position 1 CTR on AI Overview keywords falling from 0.073 in December 2023 to 0.016 in December 2025. Its earlier April 2025 study had found a 34.5% reduction; in the 2026 update, reductions for positions 2 to 10 ranged from 50.8% at position 2 to 19.4% at position 10 (Ahrefs).

Position 1 CTR on AI Overview keywords fell from 7.3% to 1.6% in two years

Click-through rate for the top-ranking page on keywords showing an AI Overview, %, desktop, global Search Console sample of 300,000 keywords (0.073 and 0.016 expressed as %)

Position 1 CTR on AI Overview keywords fell from 7.3% to 1.6% in two years

Source: Ahrefs, Update: AI Overviews Reduce Clicks by 58%, 4 Feb 2026

Seer Interactive's study of 3,119 informational queries across 42 client organisations, mainly in the US, is the only one that separates paid and organic clicks and checks whether a brand was cited in the overview. In Q3 2025, organic CTR was 1.45% with no AI Overview, 0.52% where an overview appeared and the brand was not cited, and 0.70% where it was cited. Paid CTR was 13.88% with no overview, 4.14% when not cited and 7.89% when cited (Seer Interactive, 2024). From June 2024 to September 2025, organic CTR on AI Overview queries fell 61% and paid CTR fell from 19.70% to 6.34% (the source gives the decline as both 65% and 68%); on queries without an overview, organic fell 41% and paid 32% (Seer Interactive). Cited brands saw about 35% higher organic and about 91% higher paid CTR than uncited brands, though Seer states "We cannot definitively prove that citation causes higher CTRs" and the source contains internal inconsistencies.

Paid CTR fell from 13.88% to 4.14% when an AI Overview appeared and the brand was not cited

Organic and paid click-through rate by AI Overview presence and citation, %, Q3 2025, 3,119 informational queries, 42 organisations, mainly US

Paid CTR fell from 13.88% to 4.14% when an AI Overview appeared and the brand was not cited

Source: Seer Interactive, AIO Impact on Google CTR: September 2025 Update, 4 Nov 2025

Wider traffic data point the same way. Organic search referral traffic to US sites fell from over 2.3 billion visits in July 2024 to 1.8 billion in June 2025, and Semrush measured the AI Overview share of US desktop queries rising from 6.49% in January 2025 to 20.22% in April 2025 (Digiday). BrightEdge reported search impressions up "over 49%" and click-throughs down nearly 30% in the year from May 2024, with AI Overviews in "over 11% of Google queries", without describing its method. Zero-click news searches rose from 56% to nearly 69% between May 2024 and May 2025 (Digiday).

AI assistants send some traffic back, but not much yet. ChatGPT sent 396.8 million referral visits to 1,000 tracked sites in June 2025, 81.7% of the AI referral traffic those sites received (Digiday). BrightEdge puts AI platforms at "less than 1% of referral traffic" across its clients from January to August 2025.

StudyHeadline findingGeography, period, sampleSource
Pew Research Center8% vs 15% click rate with vs without AI summary; 1% click inside summaryUS; Mar 2025; 900 adults, 68,879 searchesPew Research Center [Industry]
Ahrefs58% lower top-result CTR when AI Overview presentGlobal desktop; Dec 2023 to Dec 2025; 300,000 keywordsAhrefs [Industry]
Seer InteractiveOrganic CTR down 61%, paid from 19.70% to 6.34% on AI Overview queries; cited brands higher CTRMainly US; Jun 2024 to Sep 2025; 3,119 queriesSeer Interactive [Industry]
BrightEdgeImpressions up over 49%, click-throughs down nearly 30%; AI under 1% of referralsClient data; 2024 to 2025; method not describedBrightEdge via GlobeNewswire; BrightEdge [Industry]
Similarweb via DigidayZero-click news searches 56% to nearly 69%; ChatGPT 81.7% of AI referrals to tracked sitesGlobal/US; May 2024 to Jun 2025Digiday [Industry]
Similarweb and GraphiteChatGPT ad CTR 0.73% US, 0.74% UK, 1.91% Brazil (about one month of data); worldwide CTR up approximately 112%Nine markets; 30 Mar to 16 Sep 2026; 95,000+ US adsPPC Land [Industry]

Clicks on ChatGPT ads

The largest dataset on ChatGPT ad clicks is the Similarweb and Graphite panel, which covers nine markets from 30 March to 16 September 2026. All-time CTR ranged from 0.60% in New Zealand to 1.91% in Brazil, with the US at 0.73%; Brazil and Mexico rest on about one month of data (PPC Land). Similarweb reports worldwide CTR up approximately 112% over 30 March to 14 September 2026, without stating which weeks are compared (PPC Land). No Middle East market is in the panel (PPC Land).

ChatGPT ad CTR ranged from 0.60% to 1.91% across nine markets, with no Middle East data

All-time ad click-through rate by market, %, 30 March to 16 September 2026 (Similarweb and Graphite panel). Brazil and Mexico rest on about one month of data

ChatGPT ad CTR ranged from 0.60% to 1.91% across nine markets, with no Middle East data

Source: PPC Land on Similarweb and Graphite, 4 Oct 2026

Advertiser-reported results are scattered and come from North America. One brand tracked by Adthena saw 0.91% CTR against a 6.4% Google search benchmark in the same sector, and agency executives quoted by MediaPost reported CTR of about 0.6% (Trending Topics, Jakob Steinschaden). Common Thread Collective self-reported 0.94% CTR on USD 9,620 of spend, a single-advertiser anecdote rather than a benchmark (Search Engine Journal). On the vendor side, Criteo reports CTR about three times comparable formats, and Microsoft claims Copilot ads achieve "73% higher click-through rates" than its own search ads, on Microsoft first-party data. Neither vendor claim is independently audited.

The only UAE observations available are ONS's own two anonymised client tests, which have no published dates, spend or revenue. Recalculated from published impressions and clicks, Test A recorded 196 clicks from 4,797 impressions (196 / 4,797 = 4.09%), Test B 451 clicks from 20,663 impressions (451 / 20,663 = 2.18%), and the two combined 647 clicks from 25,460 impressions (647 / 25,460 = 2.54%). These are two selected tests, not a benchmark, and they cannot be compared with the panel figures above, which measure different populations over different periods. Chapter 11 discusses them in full.

ONS reading The click research points one way: when an AI answer sits at the top of a page, fewer people click anything, organic or paid, and being cited in the answer is associated with keeping more of the clicks that remain. For ChatGPT ads, panel CTRs below 1% in the US and UK mean an advertiser should expect to pay for attention in an environment where the answer itself satisfies many users. The commercial test is not the CTR, which varies widely by market, but the quality of the enquiries that do arrive.

What this means for UAE businesses

The mechanisms above apply in the UAE, but the evidence about their effects does not yet. AI Overviews are available to all users in the UAE and in Arabic, and AI Mode has supported Arabic since October 2025, so the answer format that the US studies examined is already in front of UAE searchers (Google Search Help; Google MENA Blog). Whether it changes UAE click behaviour by the same amount is unknown. No UAE or GCC study of AI referral traffic or AI Overview click impact was found (see Gaps).

UAE users appear receptive to AI answers. In KPMG's 2025 survey, 97% of UAE respondents used AI for work, study or personal purposes and 65% were willing to rely on information provided by an AI system. They also want transparency: 86% agree companies must ensure people can detect AI-generated content, and in YouGov's UAE sample of 1,000, 76% said it is important that content states it was made with generative AI. No survey of UAE attitudes to ads inside AI chatbots was found.

UAE law already requires the separation that OpenAI describes. Article 33(6) of Cabinet Resolution No. 68 of 2024 requires that "Advertisements shall be clearly identified, distinct, and separate from other editorial and media content" (UAE Legislation portal); Article 33(2) bars "incorrect or misleading claims"; and Article 33(9) bans direct or indirect promotion of alcohol, tobacco and smoking (UAE Legislation portal). Ads for real estate, health services, financial services, education and food require approval from the relevant authority (UAE Legislation portal). The resolution does not name channels or mention AI (UAE Legislation portal). Health advertising needs regulator approval in any channel: the MOHAP health advertisement licence covers "multiple media channels and electronic platforms in the country".

Privacy rules for ad personalisation are less settled. The UAE PDPL has been in force since 2 January 2022, but its implementing regulations "have still not been issued" as of June 2026, with no later issuance found by 28 October 2026 (Morgan Lewis; Primerus / Reina Consulting). In the DIFC, prior opt-in consent is listed for online behavioural advertising and ad personalisation (Baker McKenzie). OpenAI has not published which ChatGPT ad personalisation default applies in the UAE (Crypto Briefing via Gizmodo).

ONS reading ONS treats the UAE media law separation and approval rules as applying to ads shown inside AI assistants to UAE users, because Article 33 is written for any published advertisement. This is a planning assumption, not a legal opinion. A UAE health, real estate, finance, education or food advertiser should hold the relevant authority approval before any AI assistant campaign, in addition to meeting the platform's own policy, which outside the US generally prohibits health services and prohibits legal services.

Map the information needed at each stage

Because an assistant answers the question first and the ad sits below it, the ad and landing page must add something the answer did not (OpenAI Help). The table below is an ONS planning aid for building context hints, ad copy and landing pages around the stage of the decision.

Customer stageExample questionUseful business contentSensible next actionSource
UnderstandingWhich study route suits my schedule?Course format and entry requirementsReview programme detailsONS planning example [ONS]
ComparingWhat is included and what costs extra?Fees, inclusions and alternativesRequest a relevant quotationONS planning example [ONS]
ValidatingIs this provider suitable for my situation?Credentials, examples and limitationsSpeak to an adviserONS planning example [ONS]
ActingCan I start next month?Current availability and clear termsApply, book or purchaseONS planning example [ONS]

These are ONS planning examples, not delivery guarantees. The same individual may move between stages, compare across several websites or take action offline. A campaign should therefore support several genuine reasons to choose the business while keeping the destination focused. OpenAI's own guidance is to link to the most relevant page rather than a generic homepage. For UAE audiences, consumer protection law requires advertising and website information to be in Arabic as a minimum (Pinsent Masons).

Human judgement remains necessary

AI can help organise customer questions, draft alternative messages and find inconsistencies across a website. OpenAI now offers AI-suggested copy and images drawn from the landing page and opt-in headline translation into the user's language (Reuters via Zawya). A person who understands the service must check every material claim. Prices, professional accreditations, refund conditions, stock and appointment capacity are not details to infer from a generic model response. Under UAE law, ads must not contain misleading claims, and OpenAI reviews the advertiser, the creative, the landing page and the placement.

ONS recommends a simple working sequence: collect verified business facts, group them by customer need, draft content, have the service owner check it, then test against recorded outcomes. The value comes from faster preparation with accountable review. Producing more variations is useful only when they remain accurate and sufficiently distinct to answer a meaningful business question.

ONS outlook The formats in test point towards ads that converse rather than display: OpenAI's sponsored agents, Google's Business Agent for Leads and Microsoft's brand agents. If these reach the UAE, a business would need verified answers to its customers' common questions ready for an agent to use, and the same facts would serve organic citation. ONS's view is that preparing that fact base is worthwhile now, but no UAE launch date for any of these formats has been published and none should be assumed.

Gaps in the evidence

No public data. UAE or GCC data on AI referral traffic, AI Overview click impact, or ChatGPT ad CTR from an independent source. Searched Similarweb, Ahrefs, Seer, BrightEdge, Pew and the Similarweb and Graphite ChatGPT panel, which covers nine markets with none in the Middle East (PPC Land). A future edition should seek a UAE panel from Similarweb or a regional agency dataset with a published method.

No public data. Which ChatGPT ad personalisation default (opt-out as in the US, or opt-in as in the EU and UK) applies to UAE users; the exclusion-targeting categories in OpenAI's test; frequency capping (none documented). A future edition should read the UAE locale of OpenAI's ads help pages and privacy policy.

No public data. UAE consumer attitudes to ads inside AI chatbots, and citation patterns for Arabic-language queries. The nearest UAE data are general AI trust and disclosure preferences (KPMG Lower Gulf; YouGov, 2024).

Sources for this chapter

PUT THE INSIGHT TO WORK

Your market.
Your next move.

Explore how ONS approaches ChatGPT Ads, campaign readiness and measurement for UAE businesses.

Explore our expertise ↗