From Static Billboards to AI-Generated Context :
My 8-Year Journey Predicting Contextual Advertising
06 October 2026
The News That Validates My Thesis
OpenAI has launched a new advertising system that surfaces contextually relevant product recommendations alongside AI-generated images. When a user generates an image of a beach scene, the system suggests travel apps, resort bookings, or beach-related merchandise. The ads are not interruptions—they are context-aware recommendations embedded into the creative workflow itself.
I have been predicting this moment since 2013. Not in its exact form—I got some details wrong—but in its essence: the evolution from static, broadcast advertising to dynamic, contextually-triggered ads that respond to real-time user behavior, location, and intent.
This blog traces that journey. It's a record of what I saw coming, what I missed, and what I should have warned about more loudly.
2013: Context Arrives (But Quietly)
In September 2013, I received an update from InMobi, the world's largest independent mobile ad network at the time. I archived it, because the capability they were announcing felt significant:
"InMobi SmartAds enable rich media creatives to include dynamic feeds such as live stock prices, RSS news feeds, sports scores, or movie show times, and are influenced and triggered by behavior based on the user context (location, weather conditions and more)."
Email - "FW: InMobi Feed - Updates from Blog, News, Press" - September 24, 2013
Their example was striking:
"Strengthen a coffee chain brand by promoting hot coffee on a cold day and iced coffee on a hot day, based on local temperature conditions; and, Promote location based ads by displaying offers at a nearest store."
Email - "FW: InMobi Feed - Updates from Blog, News, Press" - September 24, 2013
This was the birth of what I would call Layer 1 Contextual Advertising: the system doesn't know who I am; it only knows the environment I'm in. It rains → inference → show umbrella ads. I'm near a restaurant → inference → show food ads.
But I could see this was only the beginning.
2018: My Vision—Context as Behavioral Fusion
In August 2018, I published a piece called "Future of Advertising." I was excited and disturbed in equal measure. I sent it directly to Rajan Anandan at Google. Here's what I wrote:
"But Ads, either on our TV screens or on a Billboard, are STATIC. It is only a matter of time before Google starts displaying on the sides of city transport buses, DYNAMIC and CONTEXTUAL advts!"
Blog - "Future of Advertising" - August 30, 2018
Then I laid out the scenario—the one that has haunted me ever since:
"Say, a spectacle-wearing pedestrian is standing on the footpath, in front of a shop of SUNAYAN OPTICIANS and sees a BEST bus approaching. He had visited an eye doctor in the morning (record on Mobile). Lo and behold! He sees on the side of the bus, an ADVT from SUNAYAN OPTICIANS, saying: FREE eye check up for the next 15 minutes! That advt turns into an ad from AJANTA CHEMIST (located 50 feet away down the road), where a pedestrian has a mobile in his hand, using which, he had ordered PENTOCID tablets in the morning. To his amazement, he notices that the Google Ad urges him to buy ANTACID tablets at 20% discount - of course, from AJANTA CHEMIST!"
Blog - "Future of Advertising" - August 30, 2018
The key insight I had in 2018: Context is not just environmental. It's behavioral. It fuses:
- Where I am (location)
- What I've done (purchase history)
- What I'm about to do (real-time intent)
And it changes dynamically, in real time.
I went even further:
"Google is developing devices (Smart phones / Smart watches / Wrist bands / Google Assistant and even contact lenses with Augmented Reality / Virtual Reality / Brain Computer Interfaces and a hundred other IoT devices surrounding you 24x7), all of which will "deliver" those Ads to your brain where these Ads will appear like "Dreams"!"
Blog - "Future of Advertising" - August 30, 2018
I was describing what I now call Layer 2 Contextual Advertising: behavioral + environmental fusion. And I was alarmed by the implications. But I didn't fully grasp where it was heading.
2022: Context Goes Mainstream (TV Edition)
By January 2022, I was seeing the principle play out across different media. TV, the last bastion of broadcast advertising, was about to go contextual. I published "Time for Getting Targeted Ads on TV?"
I opened with an observation:
"For a long time now, Social Media sites have been closely tracking / capturing our 'Clicking / Viewing' history. Then using this data to target relevant Ads to each one of us (on mobile / tablet / PC). But, as far as TV Channels are concerned, they bombard us with all kinds of Ads, at all times (carpet bombing)."
Blog - "Time for Getting Targeted Ads on TV?" - January 13, 2022
I then made a prediction:
"It seems, this may soon change and TV channels will: Target to each TV set, advts based on what is being viewed. Charge advertisers, based on 'Number of Ad-impressions delivered' (rather than time based)."
Blog - "Time for Getting Targeted Ads on TV?" - January 13, 2022
This prediction was validated faster than I expected.
2023: Addressable TV Proves the Thesis
By December 2023, Tata Play announced exactly what I had predicted. They launched addressable TV ads—the ability to show different ads to different households watching the same channel at the same time. I published "Personalizing Advertisements: As Envisaged" to reflect on how my old predictions were now shipping.
I quoted Tata Play's CEO:
"TV typically is meant for the mass advertising as the same ad is served to millions of households. But any ad that plays is not relevant to all the viewers leading to a spillover. Hence, brands do not get full bang for their buck especially when it comes to ads for premium or niche products. With addressable ad solutions, we can now split a TV ad spot into multiple beams, each beam carrying different ads for distinctive audience cohorts, in line with the requirements of advertisers."
Blog - "Personalizing Advertisements: As Envisaged" - December 11, 2023
And I reflected on my own foresight:
"5 years back, I sent my following blog as email to Google: Future of Advertising ………………………………… 30 Aug 2018"
Blog - "Personalizing Advertisements: As Envisaged" - December 11, 2023
The pattern was undeniable by 2023:
- 2013: Context = Environment (weather, location, time)
- 2018: Context = Environment + Behavior (purchase history, location, intent)
- 2023: Context = Aggregated behavior across TV viewing + demographics
But something else was coming. I could feel it. And I hadn't fully articulated the danger.
2023: The Dark Patterns Warning
In July 2023, I published "Manipulative Advertising? You Ain't Seen Nothing!"
I was responding to news that Meta was being fined in Norway for behavioral advertising without consent. But I saw a broader pattern emerging. I wrote about the risks I was observing.
And then I listed the dark patterns I was seeing:
"I hope that, in this process of delivering "targeted advts" to TV viewers, you are not resorting to any of the following "dark patterns" methods: False Urgency / Basket Sneaking / Subscription Traps / Confirm Shaming / Forced Action / Nagging / Interface Interference / Bait and Switch / Hidden Costs / Disguised Ads"
Blog - "Manipulative Advertising? You Ain't Seen Nothing!" - July 1, 2023
But I didn't have a name yet for what was coming. I didn't see that the greatest dark pattern would be seamlessness itself.
2026: The Layer I Didn't Predict
Now, in 2026, OpenAI has done something I didn't fully anticipate in 2018.
They've created Layer 3 Contextual Advertising: using the user's own generative action as the advertising signal.
In 2018, I imagined context as inference:
- I visited an eye doctor → inference → show optician ads
- I'm standing near a chemist → inference → show pharmacy ads
- I ordered antacids → inference → show antacid discount ads
But OpenAI's system uses a different model:
- I am generating a beach image → that act of creation becomes the targeting signal
- I choose to visualize a vacation scenario → OpenAI suggests resort bookings
- My creative intent is captured in real-time → ads appear based on that intent
This is generative context. The ad is not inferred from my past behavior. It's native to my present action.
Why I Got This Wrong (And Right)
What I Predicted Correctly
Context beats broadcast. I wrote in 2018 that contextual ads would outperform mass advertising. OpenAI's model proves this at massive scale. The more personalized the signal, the higher the conversion.
Context multiplexes across signals. I predicted that ads would fuse location + behavior + intent. OpenAI's model does exactly this—combining generation history, semantic intent, and user profile.
Context scales across mediums. I watched it move from mobile (InMobi, 2013) to TV (Tata Play, 2023) to generative AI (OpenAI, 2026). The principle is universal.
The endpoint is invisibility. I wrote in 2018: "these 'dreams' might one day, even deliver 'solutions' to problems." In 2026, OpenAI's ads feel like solutions, not promotions.
What I Fundamentally Missed
Generative platforms are advertising vectors themselves. I assumed contextual ads would remain a layer on top of content. I didn't foresee that AI systems would embed ads into the content generation process itself. The ad becomes part of the creative tool.
Seamlessness is the new dark pattern. In my 2023 dark patterns warning, I didn't list "seamlessness" as a danger. But it is. When an ad feels like part of the interface, my cognitive distinction between "recommendation" and "promotion" collapses.
Consent gets hijacked by design. When I use OpenAI to generate an image, I consent to "use of my generation data to improve the product." That consent is now being repurposed to justify serving targeted ads. The two are indistinguishable to me.
The Three Layers of Context (Mapped Out)
Now, looking back over 13 years, I can identify what I've been tracking:
Layer 1: Environmental Context (2013 - InMobi)
- Signal: Weather, location, time of day
- User Visibility: Clear ("It's cold, so show hot coffee ads")
- Data Required: Minimal (location + weather API)
- User Control: Moderate (I can disable location tracking)
- Ethical Risk: Low (transparent connection between context and ad)
Layer 2: Behavioral Context (2018-2023 - Google, Facebook, Tata Play)
- Signal: Purchase history, search history, viewing history, location
- User Visibility: Opaque (I don't know why the ad appeared)
- Data Required: Massive (years of behavioral tracking)
- User Control: Low (consent buried in terms of service)
- Ethical Risk: High (invisible inference from private data)
Layer 3: Generative Context (2026 - OpenAI)
- Signal: What I am actively creating in real-time
- User Visibility: Ambiguous (is this an AI suggestion or a paid ad?)
- Data Required: Deep (real-time creative state, generation history, semantic intent)
- User Control: Minimal (the recommendation is embedded in the interface)
- Ethical Risk: Critical (authority confusion between platform utility and revenue model)
Each layer is more powerful and more opaque than the last.
The New Dark Pattern: "Authority Confusion"
In my July 2023 dark patterns blog, I listed "Disguised Ads" as a concern:
"Disguised Ads: Advertisements that are designed to look like other types of content, such as news articles or user-generated content."
Blog - "Manipulative Advertising? You Ain't Seen Nothing!" - July 1, 2023
But OpenAI's ads are not disguised. They're woven. There's a critical difference.
A disguised ad pretends to be something else. A woven ad is indistinguishable from the interface itself. I don't think "this looks like an ad pretending to be content." I think "this is what the AI is suggesting to me."
I'm calling this new dark pattern "Authority Confusion."
It's the hijacking of my trust in the AI's recommendations and redirecting that trust toward commercial interests.
When OpenAI suggests a resort during image generation, I associate that suggestion with the AI's utility and intelligence. But the resort suggestion is there because OpenAI has a revenue-sharing agreement with a travel booking platform.
I confuse the AI's authority with the advertiser's influence.
What Should Happen Now
For OpenAI (and Every AI Platform)
- Label every recommendation as either utility-driven or paid. Visual separation. Verbal clarity. No ambiguity.
- Separate the teams. The team building image generation should not influence which products get recommended. Conflict of interest is not negotiable.
- Offer true opt-out. Users must be able to disable all commercial recommendations and still access all creative features without degradation.
- Publish my data sources. Which recommendations are paid? What % of recommendations are commercial? I deserve this transparency.
For Regulators
- Define "native ads" as ads. If a recommendation is paid, it's an advertisement. The medium doesn't matter. The platform doesn't get to relabel it.
- Require intent disclosure. Users must understand why a product was recommended—behavioral history, paid placement, relevance signal, or AI inference. Pick one and state it.
- Prohibit "authority hijacking." Generative systems cannot recommend products based on real-time creative inference without explicit, granular consent. "Consent to improve the product" is not sufficient.
- Mandate audit trails. Every ad recommendation must be traceable. Regulators should be able to ask: "Why was this resort shown to this user?" and get a detailed answer.
For Users
- Be skeptical of seamlessness. If an ad feels like part of the interface, it probably is—and that's by design.
- Ask whose incentive is served. OpenAI benefits from every click-through. The AI doesn't benefit. Remember the difference.
- Demand transparency before convenience. A frictionless experience is not the same as an ethical one. Insist on both.
The Broader Reflection: What I Should Have Warned About
Looking back at my 2018 blog, I see this:
"Millions may not care what Google knows about them (- and how it got to know all about them), if Google can make their lives 'less miserable / less struggle / more convenient / more comfortable / more happy'."
Blog - "Future of Advertising" - August 30, 2018
I wrote that line with some ambivalence. I wasn't endorsing it; I was observing it. But I should have made a stronger moral claim.
The fact that something is convenient does not make it ethical. And the fact that I benefit from a trade-off does not make the trade-off voluntary.
When OpenAI serves me a contextual ad during image generation, I might find it useful. That's the appeal of contextual advertising—it can be genuinely relevant. But relevance doesn't equal consent. And convenience doesn't equal autonomy.
In 2018, I predicted the mechanism of contextual advertising correctly. In 2026, I'm realizing I should have predicted the ethics more carefully.
The Arc: From Prediction to Reckoning
2013: "Context-aware ads are coming. This is clever."
2018: "Context-aware ads will fuse behavior, location, and intent. This is powerful—and alarming."
2023: "Dark patterns are proliferating. This is dangerous."
2026: "Context-aware ads have become invisible. I've lost the ability to distinguish between the AI's recommendation and the advertiser's influence. This is a moral failure."
I predicted the technology. I did not predict the acceptance. I described the mechanism but underestimated the marketing. I warned about dark patterns but didn't see that the darkest pattern would be the seamless integration of commerce into creativity itself.
The Final Question
In my 2018 blog, I asked:
"When 5G – Quantum Computing – DNA based Memory Storage – Li Fi – Face Recognition – AR – VR – AI – Blockchain etc mature, what I have envisaged above, is bound to happen."
Blog - "Future of Advertising" - August 30, 2018
It's now 2026. Those technologies have matured. My prediction came true.
But I should have asked a different question in 2018:
"Just because it's possible, does that mean it's permissible?"
The answer is no. And I wish I'd said so louder, eight years ago.
Call to Action
To OpenAI and every generative AI platform now embedding ads into creative workflows:
I have a choice. I can make contextual advertising transparent—truly, radically transparent—about who benefits and why. That's not a cost. That's a moat. It's also the right thing to do.
To Advertisers:
Don't assume that because an ad is contextual, it's ethical. Targeting someone based on what they're generating in real-time is the apex of precision. It's also the apex of intrusion. Use it responsibly. Fund transparency. Don't hide behind the phrase "AI-generated recommendations."
To Regulators:
The old definitions of "advertisement" and "recommendation" no longer map to reality. We need new categories for "generative context ads." We need audit rights. We need to act before this becomes the standard practice everywhere.
To My Readers:
Trust the AI's utility. Be skeptical of its neutrality. They're not the same thing. An AI can be brilliant at image generation and still be incentivized to serve ads. Both things are true.
With regards,
Hemen Parekh
October 2026
Sources (Contextual Advertising Series)
| # | Title | Date | About |
|---|---|---|---|
| 1 | FW: InMobi Feed - Updates from Blog, News, Press | 2013-09-24 | InMobi SmartAds launch: first context-aware (environmental) mobile ads |
| 2 | Future of Advertising | 2018-08-30 | Prediction of dynamic, contextual ads fusing behavior + location + intent |
| 3 | Time for Getting Targeted Ads on TV? | 2022-01-13 | Prediction that TV will adopt behavioral targeting like digital media |
| 4 | Personalizing Advertisements: As Envisaged | 2023-12-11 | Tata Play addressable TV launch validates 5-year-old predictions |
| 5 | Manipulative Advertising? You Ain't Seen Nothing! | 2023-07-01 | Catalog of dark patterns in modern advertising; early warning |
No comments:
Post a Comment