ChatGPT Ads Is Available in Romania: A Complete Guide for Businesses

As of 3 September 2026, self-service OpenAI Ads Manager is officially available in Romania, while the advertising platform is gaining new campaign management, audience and conversion-measurement capabilities. For businesses, this is not simply another place to buy media. ChatGPT Ads is entering a stage where campaigns can be built, monitored and optimized in ways that are closer to established digital advertising channels.

On 31 August, OpenAI announced the expansion of direct access across Europe, India, the Middle East and North Africa. Romania is listed as Available in the official Ads Manager availability page. At the same time, the weekly advertiser update introduces an Ads Manager plugin for ChatGPT and Codex, more flexible custom audiences, additional conversion-matching signals, card-level carousel reporting and new pacing for campaign-total budgets.

The essential distinction: some capabilities are already documented and active, others are rolling out gradually, while optimization that includes view-through conversions has only been announced for a later release. A business should not change its measurement setup or budget until it checks which controls are present in its own account.

What OpenAI has confirmed about ChatGPT Ads

OpenAI says ads are labeled and separated from ChatGPT answers. Advertising does not influence the assistant's answers, advertisers do not receive access to private conversations, and people receive controls over ad personalization. Whether a user sees ads depends on plan, region and the stage of the product rollout.

For advertisers, the product is no longer an impressions-only experiment. Current documentation covers CPM, CPC and oCPC objectives, geographic and platform targeting, custom audiences, product feeds, the OpenAI Pixel, Conversions API, exportable reports and self-service access. OpenAI says the platform is available in more than 40 countries, although each account, campaign and ad remains subject to eligibility, verification and applicable policies.

The advertiser updates, separated by release status

The most useful way to read this update is not as a feature list but as a separation between what can be used now and what should only be monitored. This prevents a marketing plan from relying on a control that has not reached the account yet.

The Ads Manager plugin: conversational campaign operations

OpenAI's update describes a plugin that lets an advertiser work with Ads Manager directly from ChatGPT or Codex through natural-language instructions. A website, brief or objective can become the starting point for launch-ready ad drafts. The plugin can propose copy variants, update existing structures, investigate delivery problems, review performance and prioritize recommendations.

This can reduce mechanical work, but it does not remove decisions that belong to the business. Plugins use authorized connections and remain limited by account rights, the user's role and workspace settings. OpenAI notes that availability varies by plan, region and workspace. In the advertiser workflow, recommended changes are previewed and require confirmation before they are applied.

A responsible workflow should therefore look like this:

1. Verified brief. A person establishes the objective, offer, audience, countries, budget, restrictions and landing page.
2. Generated draft. The plugin can create the structure and variations without turning its first proposal into the final decision.
3. Commercial and legal review. Claims, prices, availability, policies and usage rights for creative assets are checked.
4. Preview and confirmation. The authorized person reads exactly which entities and settings will be created or changed.
5. Post-change verification. The campaign is checked again in Ads Manager, including objective, budget, dates, countries and selected conversion.

The real value is not that “AI runs ads by itself.” It shortens the path from analysis to a reviewable draft. Responsibility for the budget, offer and data remains with the advertiser and the people who were granted access.

Custom audiences become easier to maintain

OpenAI now allows advertisers to add, remove or replace members without recreating an audience or rebuilding campaigns that use it. One file may combine supported identifiers such as email, phone, SHA-256 email or phone values, and Google Advertising ID. Audiences can exceed five million members through additional uploads.

Audience size still determines how the list can be used. Campaign inclusion and ad-group bid multipliers require at least 25,000 matched users, while OpenAI recommends at least 100,000 for a more useful inclusion or bid-adjustment signal. An audience with fewer than 25,000 matches can be used for exclusions. This is valuable when a company wants to avoid advertising to existing customers or recent purchasers even though the list is too small for targeting.

A technically accepted list is not automatically a legally usable list. The company must be able to explain the source of the data, the purpose, the notice given to people, the legal basis and the retention period. OpenAI deletes uploaded audience files after processing, usually within 24 hours, but this does not replace the advertiser's obligations for its own database and data transfer.

Conversion matching: more fields still require clean measurement

The advertiser update expands the information eligible for matching. Pixel can work with additional signals associated with phone, name, region and postal code, while Conversions API supports lists for more identifier types, including email, phone, external ID, first and last name, country, city, region, postal code and Android GAID. Current developer documentation should always be followed for formatting, normalization and hashing.

Additional fields can help when a direct click identifier is missing, but they do not repair a bad event. If a form submission is sent twice, a page view is treated as a lead, or order value is fictional, the system receives a richer signal about the wrong outcome.

OpenAI recommends using Pixel and Conversions API together for more resilient measurement. When the same action is sent from both browser and server, both copies need the same event ID for deduplication. The oppref parameter appended to a landing-page URL after an ad click should survive redirects and navigation, then be included in a server-side event when available.

The correct order: event validation, consent and deduplication come before additional matching signals. Contact data should be transmitted only when its use is permitted, clearly explained and supported by the necessary consent.

What card-level carousel reporting can reveal

In product-feed campaigns, one ad may contain several cards. The new reporting separates impressions and clicks for each card, helping a retailer see whether the first product captures nearly all attention, whether later cards rarely become viewable, or whether a product attracts visits without producing sales.

A card impression is recorded when that individual product becomes viewable in the carousel. It is different from the billable ad impression and should not be added to paid impression totals. The distinction matters because one billable ad impression may create multiple opportunities for individual product cards to be viewed.

Useful analysis does not stop at the card with the highest click count. Position, inventory, price, margin, conversion rate and revenue also need comparison. Otherwise, a visually appealing product can appear to be the winner while consuming attention without a commercial outcome.

Total budgets and view-through conversions are separate changes

OpenAI has announced more even pacing over time for campaigns using a total budget. This is a pacing rule, not a promise that every day will spend the same amount. Delivery can rise when opportunities are stronger and fall when inventory or relevance is weaker. Advertisers should monitor cumulative spend and business outcomes rather than reacting to one day's variation.

Ads Manager documentation distinguishes campaign-total and daily budgets. A total budget caps spend over the complete campaign period. A daily budget is an average target over seven days and can fluctuate within limits published by OpenAI. A daily budget may offer easier control for a first test, while a total budget gives a clearer overall ceiling to a fixed-duration campaign.

The second distinction concerns attribution. Eligible accounts can currently see separate one-day view-through conversions when a person converts after an eligible impression and no qualifying click receives credit. These conversions are supplemental. OpenAI says they do not currently affect CPA, post-click conversion rate, bidding, billing or conversion optimization.

The future model described in the advertiser update is different. It is expected to optimize using both click-through and view-through conversions, with impression-based billing. Until it launches with complete documentation, today's VTA reporting should not be assumed to change how an oCPC campaign bids. Current oCPC optimizes toward a post-click conversion and bills valid clicks.

What self-service access means for a business

Direct access and a conversational plugin do not reduce campaign setup to uploading an image and pressing a button. The configuration remains a system of decisions that must stay consistent:

1. Account and business data. The legal entity, country, currency, time zone, logo and access roles must be correct. Business or representative verification may be required.
2. Objective and measurable outcome. Awareness, clicks and conversions are different results. A validated form, qualified call or purchase should be defined before a lead campaign launches.
3. Structure and context. Campaigns and ad groups describe situations where the offer is relevant. Very broad language can dilute the match, while excessive restriction may limit delivery.
4. Targeting and budget. Geography, platforms, audiences, dates, spend limits and attribution rules must be reviewed. Media spend paid to the platform is separate from setup and management.
5. Ad and landing page. The headline, description, image and URL should promise the same thing a person finds after clicking. The page must be fast and clear on mobile.
6. Conversion and control. The selected event is tested end to end, including deduplication and reconciliation with real CRM or store data.
7. Review and optimization. Approval, delivery, cost, outcomes and lead quality are monitored after launch. Changes follow evidence, not an automated recommendation that merely sounds persuasive.

Why conversational intent is different

In conventional search, a person compresses a need into a few words. In a conversation, they can explain what they want to buy, for whom, within which budget, which conditions matter and what they have already tried. The signal is richer, but it does not guarantee that every ad will be relevant or that the person is ready to purchase.

If someone compares appointment applications for a clinic, a useful message is not a generic claim about digital transformation. It should clarify calendar integration, notifications, data protection, delivery time and implementation. In this channel, the difference between a useful and intrusive ad depends heavily on how precisely the offer answers the context.

Who should test now and who can wait

A pilot can make sense for businesses whose products or services are compared before purchase: B2B software, education, travel, professional services, technical products, property, home services or commerce with a longer decision process. In these areas, a user has questions, criteria and objections that a conversation may expose.

Companies with very low margins, incomplete tracking, slow pages or offers that are difficult to distinguish can wait. A new channel does not repair an unclear commercial proposition. Businesses in sensitive sectors also need to review platform policies, local rules, permitted claims and data handling before preparing ads.

Checklist before a first campaign in Romania

1. Define the commercial outcome. Separate purchases, quote requests, appointments and qualified calls. A visit is not enough for optimization.
2. Validate the conversion signal. Remove duplicate events, fictional values and intermediate actions incorrectly treated as final outcomes.
3. Preserve measurement identifiers. Test oppref, UTM parameters, redirects and the same event ID for an event sent through two sources.
4. Audit the audience source. Confirm data origin, consent, matched size and the intended inclusion or exclusion use.
5. Prepare the offer for real questions. Document price, duration, compatibility, guarantees, limitations and support.
6. Build a coherent landing page. Continue the ad promise with evidence, conditions and one simple next step.
7. Set a comparison baseline. Record the cost and quality of outcomes from existing channels before starting the test.
8. Prepare a controlled experiment. One budget, period, offer and stop criterion support a better decision than many simultaneous changes.

What to measure beyond the click

A minimum scorecard should include conversion rate, cost per qualified outcome, value or margin, time to conversion and lead rejection rate. Carousel card viewability and clicks can be added as diagnostic indicators, but they are not business outcomes.

Attribution requires restraint. A person may discover a solution in ChatGPT, search the brand on Google later and return directly. If every platform claims the entire conversion, the combined reports exceed reality. CRM data, campaign codes, geographic or time-based tests and branded-search analysis provide a healthier view.

OpenAI warns that conversion data may take 24–48 hours to appear. Differences from analytics do not automatically indicate an error. Attribution windows, time zone, consent, browser storage, deduplication and modeled measurement can all produce different totals.

Where SEO and promotion preparation fit

Advertising in ChatGPT does not replace organic visibility, the website or other campaigns. A user may verify the brand in Search, compare reviews and return on another device. The Romanian ChatGPT advertising through OpenAI Ads service covers commercial preparation and management, while AI and LLM optimization separately helps make public company information easier to understand, attribute and verify.

The two directions should not be confused. Ads purchase visibility in a labeled advertising surface. Organic optimization makes the entity and its content clearer to search engines and AI systems without guaranteeing a citation. Budgets and performance assessment remain separate.

A responsible first 30-day pilot

Moving a large budget abruptly from stable channels into a new product would be premature. A small, well-instrumented test can show whether incremental demand exists. If results are weak, investigate separately whether the problem is the channel, audience, message, page, bidding or measurement.

Frequently asked questions

Can I launch a ChatGPT Ads campaign myself in Romania?

Yes. On 3 September 2026, OpenAI's official list marks self-service Ads Manager as Available in Romania. Direct access does not remove advertiser verification, policies, ad approval or eligibility and delivery conditions.

Can the Ads Manager plugin change campaigns on its own?

The plugin can prepare and propose actions within the granted connection and permissions. OpenAI's announced workflow previews changes and requests confirmation before applying them. Important settings should be reviewed in Ads Manager after every meaningful change.

Can I use an audience with fewer than 25,000 matched users?

Yes, but only for exclusion. Inclusion and bid adjustments require at least 25,000 matches, and OpenAI recommends an audience of at least 100,000 for those uses.

Are carousel card impressions billed separately?

No. The update says a card impression is separate from a billable ad impression. It means the individual product became viewable and supports creative and commercial analysis.

Do view-through conversions already affect optimization?

One-day VTA reporting for eligible accounts is currently supplemental and separate from click-through conversions. OpenAI says it does not affect bidding, billing or optimization today. A model that will include view-through conversions in optimization has been announced, but it remains a future capability until rollout and documentation are complete.

Will ChatGPT Ads be cheaper than Google Ads or Meta Ads?

There is not enough comparable public evidence for Romania. Media cost alone is not decisive; cost per qualified result, incremental value and measurement quality matter.

Is installing the OpenAI Pixel enough?

No. Pixel is one component. Events must be correctly defined, tested, deduplicated, reconciled with real outcomes and implemented under applicable notice and consent obligations.

What matters now

ChatGPT Ads is becoming more relevant to businesses not only because it is expanding geographically, but because it is receiving tools for the full campaign path: preparation, audience management, conversion measurement, product analysis and spend control. None of these capabilities can turn a weak offer or incorrect data into a good campaign.

For a Romanian business, the sensible next step is not a rushed migration of budget but a controlled pilot. The advantage will belong to companies that separate active features from announced ones, keep human control over changes and compare results with revenue, margin and customer quality rather than relying only on the platform report.

Verified official sources

• OpenAI – global expansion and self-service access to ChatGPT Ads, 31 August 2026
• OpenAI Help Center – Ads Manager availability by country
• OpenAI Help Center – plugins in ChatGPT and Codex
• OpenAI Help Center – setting up custom audiences
• OpenAI Help Center – conversion measurement and matching
• OpenAI Help Center – metrics, reporting and view-through conversions
• OpenAI Help Center – campaign objectives and budgets
• OpenAI Help Center – current oCPC campaign behavior

This article was updated and verified on 3 September 2026, including against the official communication sent to advertisers. The platform is in beta, and access, capabilities and rules can vary by account or change over time. Official documentation and the controls visible in the account remain the final references before launch.