51% of AI answers in the US now carry sponsored content. Turkey targeting is already open on the Google side. Check where you stand
People no longer type "which one is best?" into a search box. They ask an AI, and the answer arrives whole. Directly beneath it sits an ad slot, and that slot is now for sale. Webtures places your brand inside AI answers on both the organic and the paid side.
We have carried sixteen years of search expertise into the ad infrastructure that generative platforms are building right now: account architecture, context hint design, product feeds, creative and conversion measurement, run by one team.
GEA is to generative platforms what SEA is to classic search engines. In plain terms: advertising inside AI answers.
Generative Engine Advertising (GEA) is the practice of buying clearly labelled paid placements that sit inside, or immediately below, an answer generated by an AI system.
In classic search advertising the user types a query, sees ten blue links, and weighs the ads among them. Generative engines have no such list. The user asks a question and receives a single, finished answer. The ad is not an item competing with that answer; it is a separate card that completes it. That structural difference rewrites the rules of the auction, the targeting and the creative.
Nobody types "CRM for healthcare". They describe their company size, their sector, their compliance constraints, their budget and their deadline in a single paragraph. The ad system reads that whole paragraph. The unit of targeting is no longer the word; it is the context.
The auction multiplies every bid by a relevance score before it ranks anything. A brand that bids less but overlaps more closely with the conversation outranks a higher bidder and pays less for the privilege. That turns a budget contest into a contest of content and positioning.
Ads run in a separate system. The model does not know an ad will appear, and advertisers cannot edit what it says. You cannot buy an AI's organic recommendation of your brand. That is precisely why GEA is never enough on its own and has to be designed alongside GEO.
Two zones of the same screen. GEO gets your brand cited inside the answer as a source. GEA buys the sponsored slot directly beneath it. One is earned, the other is bought.
| Dimension | GEO · organic AI optimisation | GEA · paid AI advertising |
|---|---|---|
| Placement | Inside the answer, as a cited source and brand mention | Below the answer, as a separate card labelled "Sponsored" |
| How you get it | Earned through content, authority and technical groundwork | Bought with budget, bids and creative |
| Shapes the answer | Yes, the model synthesises your content | No, it is fully independent of the answer |
| Time to first result | 4 to 12 weeks | Hours after approval |
| Cost shape | Fixed production and infrastructure investment | Variable cost per click or impression |
| When budget stops | The gain persists and compounds | Visibility ends immediately |
| Core metric | Citation rate, share of voice, brand mentions | Impressions, clicks, conversions, ROAS |
| Level of control | Indirect | Direct |
The GEA relevance score is computed from four inputs: the conversation descriptions, the landing page, the headline and the body copy. Because the landing page sits inside that equation, structured pages produced by GEO work lower your cost per click directly. Splitting GEO and GEA across separate teams is a measurable loss of money.
GEM is the umbrella discipline covering every marketing activity on generative AI platforms. It is the counterpart of SEM in the classic world, and it stands on two legs.
The vocabulary has not fully settled in the industry yet. Webtures adopts the umbrella definition that dominates the international academic literature and marketing technology documentation: GEM is a brand's entire visibility strategy across the AI ecosystem, and GEA is the paid leg of that strategy.
◆ Umbrella discipline
GEM · Generative Engine Marketing
The whole AI visibility strategy
Organic visibility
GEO / AEO / LLMO
Paid visibility
GEA
GEM = GEO (organic) + GEA (paid)
Four structural shifts landed at once: behaviour moved, inventory filled up, the traffic turned out to be unusually valuable — and measurement systems were left behind.
"Which accounting software is the most reliable in Istanbul?" no longer goes into a search box. It goes straight to an assistant, and the reply is not ten links but one shortlist. If you are not on that shortlist, you never even reach the comparison stage.
Turkey sits ahead of the global average on this shift: it leads the world on ChatGPT's share of AI-referred web traffic, roughly 14 points above the global mean. Turkish consumers moved to conversational search faster than almost anyone else.
Advertising on AI platforms sat in the "someday" category for years. The first half of 2026 changed that completely:
Volume is still a small slice of the total. The quality picture is another matter entirely:
Why? Because this user is not browsing, they are deciding. They have described the problem paragraph by paragraph and are waiting for a shortlist. That density of intent is rare in digital marketing.
This is the industry's biggest blind spot. Most AI-referred traffic lands in standard analytics setups as "direct traffic". Paid accounts and some deep-research modes pass no referrer at all.
The result: most brands never measure the value coming from this channel, so they never fund it. Without a proper channel group, server-side tagging and a conversions API, reporting real GEA performance is simply not possible. That infrastructure is where the Webtures engagement starts.
Six modules, one operation. Every module has a concrete deliverable, and the order matters: no campaign goes live before the measurement layer exists.
◆ What we do
Why it mattersYour ad card is read immediately below the model's answer. If the model already says "every product in this category offers integrations", an integration-led headline is dead before it ships. Creative strategy is born out of this analysis.
→ Deliverable: AI visibility scorecard and opportunity map
◆ What we do
◆ Example: a weak description next to a strong one
✗ Weak
CRM software for the healthcare sector
✓ Strong
Operations and revenue leaders at health tech companies of 30 to 150 people, unhappy with their current CRM, looking for patient messaging integrations and a go-live window under 30 days
Why the gap is so wideThe weak description matches thousands of conversations a day and loses nearly all of them. The strong one matches thirty and wins all thirty, because nothing else comes close on relevance.
→ Deliverable: context hint library and ad group architecture
◆ What we do
Our positionThe expensive advertiser on this channel is not the one who spends the most; it is the irrelevant one. Relevance score is always our first optimisation lever.
◆ What we do
The logic of the formatThe user has already received a good answer and is satisfied. If your ad adds nothing the answer left out, it does not get read. Category language loses here; a concrete difference wins. Not "discover our AI platform", but "98.7% accurate transcription".
→ In depth: ChatGPT Ads management service
◆ What we do
▲ WarningWithout a custom channel group, most AI-referred traffic is reported as "direct". Any GEA reporting built before that infrastructure exists will systematically understate real performance.
◆ What we do
◆ The metric layers we report
| Layer | Metrics |
|---|---|
| Inventory | Prompt coverage, impression share, category penetration |
| Efficiency | Cost per click, cost per mille, effective cost per acquisition |
| Quality | Click-through rate, session duration, bounce, pages per session |
| Outcome | Conversion rate, average order value, ROAS, pipeline contribution |
| Synergy | Relationship between organic citation rate and cost per click |
AI advertising is not one platform. Inventory, buying method and maturity differ on every surface. We refresh this table monthly.
| Platform | Paid inventory | How you buy it | Turkey targeting |
|---|---|---|---|
| ChatGPT | ✓ Available | Its own self-serve ads manager | Not open yet |
| Google AI Overviews (top and bottom) | ✓ Available | Performance Max, AI Max, Shopping, broad match | ✓ Open |
| Google AI Overviews (in-answer) | ✓ Available | Same campaign types | US first |
| Google AI Mode | ✓ Available | Performance Max and AI Max | Expanding |
| Microsoft Copilot | ✓ Available | Microsoft Advertising, Performance Max and Showroom | Limited |
| Perplexity | · None | Not applicable | Not applicable |
| Claude | · None | Not applicable | Not applicable |
| Gemini app | · None | Not applicable | Not applicable |
The only standalone platform where you can genuinely buy an AI ad today. Ads are shown to free and entry-tier subscribers; higher tiers and under-18 users are excluded. With the minimum spend requirement gone, you can enter on a test budget.
Google never packaged its AI surfaces as a separate ad product. AI Overviews and AI Mode placements are open only to certain automated campaign types; standard search campaigns cannot reach that inventory at all. You also cannot bid on the placements directly. The one lever you control is the quality of the signal you feed the system: feed accuracy, asset depth and landing page fit.
The critical operational detail is the auto-upgrade schedule. Campaigns using automatically created assets and campaign-level broad match are being migrated to the newer campaign type in waves. Letting that transition happen unsupervised produces unpredictable swings in account performance.
The ad appears as a distinct block beneath the answer, and Copilot explains in its own words why it is showing it. Matching looks at the whole session rather than the last message. Its real differentiator: it is the only major ad platform offering targeting on professional profile signals such as company, job function, industry and seniority, a layer no other channel gives B2B and enterprise sales teams.
One detail worth flagging: a significant share of brands running Performance Max on Microsoft are already advertising on Copilot without realising it. It is one of the most common findings in our audits.
Perplexity ended its advertising programme on the grounds that sponsored content undermines citation trust, and moved to a subscription-first model. Claude has no public advertising programme either. Visibility on these platforms is won through GEO and AEO only.
Any proposal claiming you can buy ads on these two platforms is working from outdated information.
The Webtures GEM methodology runs a six-phase roadmap. The sequence is not arbitrary: each phase produces the input for the next.
Which questions cite you, on which claims, and which questions leave you out entirely. The same scan maps where ad inventory exists.
Conversation descriptions are written in the language buyers actually use. That library feeds both the paid targeting and the organic content plan.
Measurement architecture, product feed, landing page matching and technical accessibility all get configured here. Skip this phase and the campaign runs, but its results cannot be reported.
A controlled budget, tested simultaneously across several conversation themes. The goal is not to spend; it is to learn which intent cluster converts.
Findings drive narrower descriptions, landing pages split by theme, and creative rewritten around the truncation window. Raising the bid is the last step of this cycle, never the first.
Winning themes get funded, then move into the organic content plan — because as organic strength grows, landing page quality rises, the relevance score climbs and paid cost falls. That is the compounding mechanism at the heart of GEM.
AI ad inventory is not open in every market or every sector yet. We therefore ship the service in three configurations, chosen by your target market.
Cross-border ecommerce, SaaS and software companies selling internationally, inbound travel and hospitality brands, design and manufacturing firms growing in export markets.
For these brands ChatGPT advertising is available today. With no minimum spend, you can enter on a test budget. This window, before competition matures, is the cheapest position you will ever buy in this channel.
◆ Recommended · GEA PilotEcommerce, retail, service and B2B brands selling into the domestic market.
ChatGPT advertising is not open to Turkey targeting yet, but Google AI Overviews placements are available in more than 200 markets, Turkey included. The strategy has three legs: campaign architecture built for Google's AI surfaces, organic visibility through GEO, and readiness to launch on day one when ChatGPT opens the market.
◆ Recommended · GEA Readiness and GEM Full ScopeOrganisations operating in several markets that want organic and paid visibility managed as one strategy.
For these brands, splitting GEO and GEA across separate teams is a measurable loss of money. Because landing page quality feeds directly into paid cost, combining the two disciplines under one operation pays off both operationally and financially.
◆ Recommended · GEM Full ScopeBeing straight about this matters more than selling the channel. Under current platform policy the following categories are closed to AI ad inventory or face serious restrictions:
Beyond that, healthcare, financial services and legal services are broadly out of scope in markets outside the US. For health tourism, aesthetics, finance and law brands based in Turkey, that means ChatGPT advertising is not usable today.
Our recommendation is simplePut the budget into GEO and AEO instead. Being cited organically in AI answers is the only accessible route in these sectors, and also the most credible one.
New rules affecting digital advertising in Turkey came into force on 1 August 2026. Amendments to the Commercial Advertising and Unfair Commercial Practices Regulation place targeted advertising, influencer posts and AI-generated advertising inside a single transparency framework.
Where an ad uses AI characters that are hard to tell apart from real people, that fact must be disclosed clearly, comprehensibly and distinguishably.
Ads that give the impression a real person's digital replica has personally used or endorsed a product are prohibited.
Enforcement does not reduce to "was AI used?". What decides the case is the perception of person, experience and endorsement the asset creates in the consumer.
Within the Webtures GEA engagement, every creative we produce is reviewed against this framework before it goes live. For campaigns using AI-generated imagery or video, that compliance check is an inseparable part of the process.
Every answer is written to stand on its own out of context — for you and for answer engines alike.
There is no single console for advertising on AI platforms; each one has its own buying route.
For ChatGPT you register in its own ads manager and build a three-tier structure: objective, budget and country targeting at campaign level; conversation descriptions and bids at ad group level; headline, body, image and landing page at ad level. Targeting is not a keyword list but conversation descriptions written in natural language.
Google has no separate AI ad platform. AI Overviews and AI Mode placements are served through automated campaign types such as Performance Max and AI Max for Search; standard search campaigns cannot enter that inventory. Microsoft Copilot placements likewise run through Performance Max in Microsoft Advertising.
Perplexity and Claude carry no purchasable ad inventory; visibility there is earned through organic optimisation alone.
Measurement is built in two layers: in-platform reporting and your own analytics stack.
The platform reports impressions, clicks, spend, click-through rate, average cost per click and per mille, and conversions. Conversion measurement uses a pixel plus a server-side conversions API; because UTM parameters survive the click, campaigns can be tracked in your own analytics alongside every other paid channel.
The critical point: most AI-referred traffic is misclassified as "direct" in default analytics setups, and some platforms and mobile apps pass no referrer at all. Reporting built without a custom channel group that catches AI source domains, a consistent UTM taxonomy and server-side tagging will systematically understate the channel's contribution. This infrastructure is where the Webtures engagement begins.
Partly. It depends on the platform.
ChatGPT advertising is live in a limited set of markets and Turkey targeting is not yet possible. However, Turkey-based brands selling into those markets — cross-border ecommerce, software and SaaS companies selling abroad, inbound travel brands — can use this inventory today.
For Turkey-targeted campaigns the Google side is open. AI Overviews placements above and below the answer are available in Turkey just as in the 200-plus markets where AI Overviews runs, and are bought through the eligible campaign types.
For Turkey-focused brands we recommend a three-legged approach: the right campaign architecture for Google's AI surfaces, organic visibility through GEO, and readiness to launch on day one when ChatGPT opens the market.
GEM (Generative Engine Marketing) is the umbrella discipline covering all marketing activity on generative AI platforms. GEA (Generative Engine Advertising) is the paid leg of that discipline.
The cleanest way to hold the relationship in your head is the classic parallel: as SEM covers SEO and SEA, GEM covers GEO and GEA. A brand can do GEA alone, or GEO alone; GEM means running both together so that each strengthens the other.
A context hint is the targeting unit that replaces the keyword on AI ad platforms. Defined at ad group level, it is a short piece of natural language describing the conversations where your product or service is relevant.
It differs from a keyword in three ways: it is not an exact match, it does not guarantee delivery, and it describes not a word the user typed but the situation they are in. Since the user hands the model their company size, sector, constraints and timeline in a single paragraph, the targeting has to be rich enough to represent that whole paragraph.
The most effective descriptions carry three layers: audience (who they are), intent (what they are trying to do right now) and topic (the category, subcategory and the constraint that points at you).
No. Platforms protect that separation at the level of product principle: ads run in a system apart from the answer, the model does not know an ad will be shown, and advertisers cannot change the answer's content, ranking or sources. Ads are explicitly labelled and visually separated.
The practical consequence: an AI's organic recommendation of your brand cannot be bought. Being named inside the answer is earned through content quality, authority and technical accessibility. Paid visibility does not substitute for organic visibility; it works alongside it.
Cost comes out of the auction, and three buying models exist: cost per mille, cost per click and cost per action.
But the factor that actually sets your price on this channel is not the bid. The auction multiplies bids by a relevance score before ranking, and that score is computed from the conversation descriptions, the landing page, the headline and the body copy. A brand bidding less but overlapping strongly with the conversation can beat a higher bidder and pay less.
What we see in practice: in the large majority of accounts running above expected cost, the problem is not budget but description quality. That is why raising the bid is the last step in our optimisation cycle, not the first.
With the minimum spend requirement removed, the entry barrier is no longer budget but quality of preparation. For the pilot we recommend a controlled budget large enough to test several conversation themes at once; the goal is not to spend but to learn which intent cluster converts.
The right order is: measurement infrastructure first, then the description library, then landing page matching, and budget last. Skip that order and the money buys impressions rather than data.
On the paid side, ads go live within hours of clearing review and the first data lands in the same week. A meaningful optimisation decision, though, needs enough click volume per variant to accumulate, which usually takes 3 to 6 weeks.
On the organic side, being cited as a source in AI answers takes 4 to 12 weeks. Run the two disciplines together and the paid side produces fast signal while the organic side converts that signal into durable, zero-cost visibility.
Not directly. These platforms offer no public ad library of the kind social networks and search engines provide. That does not make competitor analysis impossible, but it does make it a process rather than a lookup.
Our method has three layers: repeating the conversations your buyers have, at regular intervals and under controlled conditions; reading competitors' messaging on other channels as a proxy; and above all, tracking what the AI itself says about your category. Since your ad is always read in the shadow of the model's answer, knowing what that answer says is the core input to creative strategy.
The two are not alternatives but complements, and there is a measurable technical link between them.
The GEA relevance score is computed from four inputs, one of which is the landing page. Structured pages produced by GEO work, which answer the question directly, raise that score. A higher score means a lower cost per click. In other words, your GEO investment lowers your GEA costs.
The benefit runs the other way too. GEO's hardest problem is measurement, because most organic AI traffic is invisible in analytics. GEA, with its pixel and conversions API, produces clean data. Running a paid test on the same conversation themes gives you the fastest available signal about their organic potential.
You can measure the answer instead of guessing at it. The Webtures GEA team maps where your brand and your competitors stand inside AI answers, shows which conversations carry ad inventory, and hands you a roadmap built for your market.
No cost · No commitment
Or just talk to us directly
hello@webtures.com
This page is a hub — every sub-discipline on the paid and organic side branches out from here.