The relationship between media buyers and artificial intelligence is entering a more concrete phase. It is no longer limited to the possibility of using ChatGPT to write a few copy variations or generate creative ideas. It is progressively extending to data analysis, campaign configuration, budget allocation, and, potentially, even the purchasing of advertising space.
This reflection stems from an email that has circulated in recent hours, apparently linked to an application called “MetaAI Ads GPT,” presented as a tool capable of simplifying campaign creation and management through artificial intelligence. The message, sent from the address accountdesk@reachlyne.co, contains several elements suggesting that it is not an official communication from OpenAI, Meta, or any other recognized advertising platform.
However, the alleged application itself is not the most interesting aspect. What matters is that the proposition immediately sounds credible, at least conceptually, because it describes exactly the direction in which the market is moving: an environment where advertisers no longer interact exclusively with a technical dashboard, but instead communicate with an AI system capable of reading an account, interpreting performance, and translating a business objective into a series of advertising operations.
OpenAI has already begun developing its advertising activity within ChatGPT, introducing ways to purchase and manage campaigns and developing a dedicated platform for advertisers. At the same time, Meta continues to expand Advantage+, while Google is moving AI Max beyond the beta phase and plans to progressively update some of the long-standing features of Search campaigns. We are therefore not dealing with a simple trend surrounding chatbots, but with a structural transformation in advertising intermediation.
Media Buyers and Artificial Intelligence: The Change Did Not Start with ChatGPT
When discussing AI applied to advertising, attention tends to focus on language models such as ChatGPT and Claude. In reality, media buying has long been governed by automated systems that already make a significant portion of distribution decisions.
Platforms determine who sees an ad, how much to bid in a particular auction, which creative to prioritize, and when to increase or decrease exposure. The transition from predominantly manual campaigns to systems such as Advantage+, Performance Max, and AI Max did not suddenly introduce automation. Instead, it progressively reduced the number of operational decisions directly controlled by advertisers.
Meta defines Advantage+ as a set of solutions powered by artificial intelligence and automation that can be used throughout the entire campaign process or for individual tasks. Google, meanwhile, describes AI Max as a suite capable of expanding search matching, adapting ads, and optimizing messages and destinations in real time.
This means that the media buyer’s job had already begun to change before the widespread adoption of generative chatbots. The difference is that ChatGPT, Claude, and AI agents add a new layer by making automation accessible through natural language.
Instead of entering a dashboard, applying filters, exporting data, and manually building comparisons, professionals can ask a system to identify campaigns whose performance is deteriorating, compare acquisition costs across different periods, or determine whether a decline in ROAS is caused by traffic, conversion rates, or average order value.
This conversational interface does not necessarily replace the advertising platform, but it changes how professionals interact with it.
This dynamic is consistent with what has already been observed in the transformation brought about by Meta Andromeda and the evolving role of media buyers: as algorithms assume greater control over delivery, human value shifts away from manually adjusting campaigns and toward the quality of inputs, analysis, and subsequent decisions.
From Advertising Dashboards to Conversational Interfaces
The most plausible evolution is not the immediate disappearance of Ads Manager or Google Ads, but their gradual transformation into underlying infrastructure.
Media buyers will continue to use these platforms, but they may increasingly do so through a conversational layer. They could ask a system to create a campaign for a particular service, define the maximum budget, specify geographical areas to exclude, upload creatives, and establish the desired cost per acquisition. The system would then translate those instructions into operational settings.
At a later stage, AI could automatically compare campaign data with information from the CRM, website, ecommerce platform, and management systems, providing an analysis that goes beyond the leads or purchases recorded by the advertising platform itself.
This perspective is important because one of the historical limitations of media buying stems from fragmented information. Meta knows what happens within its own ecosystem, Google interprets signals collected through its tools, while the CRM stores actual commercial data. Professionals are often forced to connect these layers manually to determine whether a campaign is generating real customers or merely conversions that appear cost-effective on the surface.
An AI assistant connected to all these sources could, for example, highlight that one campaign generates leads at a lower cost but with a significantly lower closing rate, or that a particular audience purchases less frequently but generates customers with greater long-term value.
The advantage would therefore not simply be faster execution. It would be the ability to evaluate campaigns through a more comprehensive economic perspective.
What ChatGPT or Claude Can Do for a Media Buyer Today
The most mature use of artificial intelligence in advertising does not consist of asking a model to “create a winning campaign.” A request of this kind inevitably produces generic recommendations because the system lacks the context required to make a reliable decision.
The real value emerges when the model receives well-defined data, constraints, historical information, and objectives.
It can analyze exports from Meta Ads or Google Ads, summarize large amounts of information, identify unusual variations, compare segments, classify creatives, and help formulate hypotheses to test. It can also accelerate report production by transforming complex tables into clear explanations for clients or company management.
Through connectors, APIs, and MCP-based infrastructures, models such as Claude and ChatGPT can be connected to external sources and, in some cases, to tools that read or manage advertising data. However, some of the solutions currently available are developed by third-party providers and do not necessarily represent official integrations with the advertising platforms themselves. It is therefore essential to verify who manages the connection, what permissions are required, and where the data is processed.
Data Analysis Becomes Faster, Not Automatically More Accurate
A language model can recognize that the cost per result has increased while CTR has remained stable and the landing page conversion rate has declined. It may therefore suggest that the problem occurs after the click rather than within the campaign itself.
This analysis can be useful, but it remains dependent on the quality of the data provided. Incomplete tracking, duplicated conversions, different attribution windows, or incorrectly configured events can lead the system to construct a formally coherent analysis based on unreliable information.
Artificial intelligence does not solve this problem. On the contrary, because AI can explain concepts in an organized and seemingly confident way, it can make an incorrect interpretation appear highly credible.
For this reason, Meta campaign tracking and measurement quality become even more important. The more decisions are automated, the more quickly an error at the source can spread across budgets, creatives, and distribution.
Creative Production Increases, but So Does the Noise
ChatGPT, Claude, and multimodal models make it possible to produce communication angles, headlines, scripts, images, and ad variations much more quickly. Platforms can then automatically combine these assets and distribute them according to the signals they collect.
The operational advantage is clear: a team can test far more hypotheses than it could in the past. The risk, however, is confusing the number of variations with the quality of the creative strategy.
If everyone uses the same models, the same prompts, and the same persuasive structures, the result is not necessarily greater differentiation. The opposite may happen: an exponential increase in content that is technically correct but increasingly similar.
This is already visible in the management of Meta Ads creatives, where the ability to produce numerous variations must be accompanied by a system capable of genuinely distinguishing between concepts, promises, objections, formats, and different levels of audience awareness.
Although AI can automatically multiply the number of assets, it cannot independently determine which positioning will make an offer more relevant than its alternatives.
The Media Buyer Is Not Disappearing, but Some of the Tasks That Defined the Role Are
Part of the operational workload will inevitably be absorbed by advertising platforms and AI assistants. Manual adjustments to certain bids, the repetitive creation of variations, report exports, and part of the daily monitoring process will progressively become automated.
This does not mean that every media buyer will be replaced. However, some skills that were considered essential only a few years ago will have less economic value.
Knowing exactly where to click to create a campaign, remembering the location of every setting, or being extremely fast at exporting data will no longer be sufficient competitive advantages. When a system can perform these tasks through a command, professional differentiation must emerge both before and after execution.
Before execution, value lies in the ability to select the right objective, organize data, formulate a hypothesis, and translate a business model into signals the algorithm can understand.
After execution, it lies in the ability to interpret results, recognize the limitations of attribution, distinguish correlation from causation, and decide whether to accept or correct an automated recommendation.
The change is similar to what has already happened with the spread of automated campaigns. As the comparison between Meta Advantage and manual campaigns demonstrates, the challenge is not to choose ideologically between human control and automation, but to understand which decisions should be delegated to the system and which should remain under human supervision.
The New Competitive Advantage Will Be the Quality of Context
An AI system does not automatically know a company’s margins, sales capacity, average closing times, seasonality, the geographical areas it can realistically serve, or the type of customer the business wants to avoid.
If it is given only a lead generation objective, it will attempt to generate the highest possible number of leads based on the available signals. It cannot know that the sales team can handle no more than fifty enquiries per week, that certain geographical areas generate unprofitable appointments, or that a particular service has margins too low to support the advertising cost.
The media buyer of the future will therefore need to become a designer of the context in which automation operates.
They will need to define which data enters the system, which conversions genuinely have value, which constraints cannot be exceeded, and which metrics should take priority. Their work will become less focused on continuously modifying campaigns and more centered around orchestrating data, creative, offers, sales, and technology.
The same applies to Google, where the progressive expansion of AI Max in Search campaigns shifts some control away from traditional keywords and toward the system’s ability to interpret pages, assets, queries, and conversion signals.
When platforms automate matching, bidding, and part of the ad generation process, the quality of landing pages, feeds, conversions, and information architecture becomes increasingly important. The algorithm can only optimize what it is able to read.
The Impact on the Agency and Freelance Market
The impact of artificial intelligence will not be uniform. The activities most exposed to automation are those built primarily around standardized execution: campaign configuration, descriptive reporting, generic copy production, and optimizations carried out according to repetitive rules.
These services will not disappear entirely, but they will become more difficult to sell as high-value activities. A client may reasonably question why they should pay for many hours of work to produce a report that a model can generate in minutes, or for a campaign structure that the platform itself can create through a guided process.
Agencies will therefore need to demonstrate value in areas more closely connected to economic outcomes: offer strategy, measurement design, test organization, sales data analysis, supervision of automated systems, and integration between marketing and the sales process.
Pricing models may also change. It will become more difficult to justify fees based exclusively on hours worked when automation reduces the time required for many operations. Models based on project complexity, responsibility, infrastructure quality, or economic objectives may become increasingly relevant.
At the same time, AI lowers the barrier to entry into advertising. Small businesses and professionals who previously lacked the skills to set up a campaign will be able to get started more easily. This will expand the market, but it will also increase the number of advertisers and, consequently, competitive pressure within advertising auctions.
Automation does not necessarily make advertising cheaper. If it enables more companies to enter the system, it can increase demand for attention and make the ability to build differentiated offers and creatives even more important.
What Still Needs to Improve Before AI Can Truly Manage a Campaign
The direction is clear, but the current generation of tools still presents significant limitations.
The first is transparency. Automated platforms are making an increasing number of decisions but often provide insufficient explanations for why particular distribution choices were made. A recommendation may be correct, but professionals need to understand which signals it is based on and which variables were considered.
Research into explainable artificial intelligence applied to advertising highlights precisely this issue of opacity: systems capable of optimizing targeting, pricing, and relevance have access to data that advertisers themselves cannot see, making automated decisions more difficult to evaluate and challenge.
The second limitation concerns operational reliability. An assistant connected to an advertising account should not simply execute a command. It should also recognize when a request is inconsistent, risky, or based on insufficient data. Before increasing a budget, it should verify tracking stability, product availability, sales capacity, and the actual value of conversions.
The third issue concerns permissions. Allowing AI to read data is one thing. Allowing it to modify campaigns, budgets, or creatives is another. More mature systems will need to introduce granular access levels, operation logs, approval procedures, and rollback capabilities.
The fourth relates to brand management. Platforms are introducing automated image generation and creative variations, but the results can alter products, tone, and messaging. Recent reports have shown that some companies have encountered problems with automatically generated or modified assets, making careful human oversight increasingly necessary.
Finally, the ability to integrate short-term and long-term results must improve. Platforms are highly efficient when optimizing measurable conversions, but a business does not survive on clicks, leads, or immediate purchases alone. Reputation, positioning, customer quality, profitability, repeat purchases, and the overall value of the customer relationship also matter.
Until these elements are accurately represented in the available data, automation will continue to optimize only a partial representation of the business.
ChatGPT Ads Also Opens a New Market for Advertisers
The relationship between ChatGPT and advertising is not limited to using AI to manage campaigns on Meta or Google. OpenAI is developing its own advertising ecosystem based on the idea of reaching users while they explore possibilities, compare alternatives, and make decisions.
This approach could create a form of advertising different from traditional search. On Google, users often express intent through a query. In ChatGPT, they build a conversation in which they progressively specify needs, constraints, preferences, and decision-making criteria.
For advertisers, relevance may therefore depend less on matching a single keyword and more on the ability to appear within a complex decision-making context. OpenAI has stated its intention to develop new purchasing methods, CPC systems, and measurement tools within its advertising platform.
This raises questions that remain unresolved. It will be necessary to understand how clearly advertising will be separated from organic responses, how transparency will be managed, what data advertisers will be able to use, and how conversions generated through long conversations distributed over time will be attributed.
For media buyers, this means preparing for an environment in which advertising will no longer appear exclusively in a feed or alongside a search query, but may become part of an advisory process conducted by a machine.
The Future of Media Buyers Is Not Becoming Prompt Specialists
The simplest mistake would be to replace one technical skill with another and assume that the professional of the future simply needs to know how to write prompts.
A prompt is only an interface. It can improve the quality of an output, but it cannot replace the ability to understand which problem actually needs to be solved.
A competent media buyer will not be valuable because they know a particularly sophisticated formula to enter into ChatGPT. They will be valuable because they know which data to provide, which information to exclude, which checks to perform, and how to evaluate the response they receive.
The real specialization will not be knowing how to talk to AI, but building systems in which AI can operate without drifting away from the company’s objectives.
For freelancers, this means using automation to spend less time on repetitive activities and more time understanding their clients. For agencies, it means integrating advertising data, CRM systems, creative production, and sales into a coherent process. For businesses, it means not delegating the definition of objectives to the platform, but instead providing it with better signals.
The Profession Changes When Execution Is No Longer the Main Source of Value
The email received today most likely does not represent the credible launch of an official tool. However, it accurately reflects the landscape emerging around advertising: a system in which an advertiser simply connects an account, describes the desired outcome, and allows artificial intelligence to handle the rest.
Less complexity in manual campaign configuration means greater complexity in ensuring data quality. Less time spent producing reports means greater responsibility in interpreting them. Less direct control over targeting and bidding means greater attention to the inputs that guide algorithms. Creative production becomes faster, but standing out becomes more difficult.
Media buyers whose value is based exclusively on operating an advertising dashboard will become increasingly exposed to automation. Those who understand the business, connect different information sources, supervise models, and transform data into decisions will continue to play a central role.
Not because artificial intelligence is incapable of managing campaigns, but because a campaign is not the same thing as the business it is supposed to support.
