# The attribution gap: who gets the credit in an AI-mediated world?

Published: 2026-11-17

Canonical URL: https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/

---

# **The attribution gap: who gets the credit in an AI-mediated world?**

AI is increasingly the thing standing between a brand and its audience. Five experts on what gets lost in translation - and who’s supposed to get the credit.

**With contributions from**

*Spark AI, BerylliumIV, Havas Media Network UK, Publicis Media, Unbound IA*

---

![A bunch of colorful speech bubbles on a dark blue background. The speech bubbles are arranged in a pattern, with each one having a unique design. The colors of the speech bubbles range from bright blues and greens to warm oranges and yellows, creating a vibrant and inviting atmosphere.](https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/assets/lymQU7LZuP/istock-2178906768-1208x869.jpg)

## **The attribution gap: who gets the credit in an AI-mediated world?**

AI is increasingly the thing standing between a brand and its audience. Five experts on what gets lost in translation - and who’s supposed to get the credit.

**With contributions from**

*Spark AI, BerylliumIV, Havas Media Network UK, Publicis Media, Unbound IA*

---

![](https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/assets/5yAnqwTrhu/istock-2178906768-1208x869.jpg)

You’ll remember a version of this from your school days: someone mutters a barely audible funny line under their breath and a second later the class clown repeats it, but louder - getting the laugh. The joke was never theirs, but timing and volume are what the class remembers, not the origin.

In some ways marketing measurement has quietly worked the same way for years, sometimes crediting whichever touchpoint is clearest and nearest to the sale, ignoring whatever built the idea behind it. AI hasn’t created that problem. But it has made it much harder to ignore. Increasingly there’s no “loud kid” to point to either. The source disappears, the influence disappears, and the question of who actually did the work becomes almost impossible to answer with a straight face.

So, who gets the credit when AI is the one doing the repeating? The honest answer, depending on who you ask, is: it depends what you mean by credit - and whether chasing it is even the right fight.

## **The invisible interaction**

![](https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/assets/wv2y5YJE7q/istock-1164708053-1290x813.jpg)

![Emma Jackson](https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/assets/u2Wc8PJQrB/emma-jackson.png)

**Emma Jackson**  
*Executive producer* working in frontier technology, and an *AI transformation Coach* for Spark AI

Perhaps the best place to start is with what’s disappearing inside the output itself. Emma Jackson, an executive producer working in frontier technology, and an AI transformation Coach for Spark AI, splits “who gets the credit” into two separate questions: where did this come from, and who made it?

1. **Provenance: “Where did this come from?”**   
“As users,” Jackson says, “we can’t see what in a model’s training shaped its answer. But that doesn’t make us powerless. We can record the sources we provide or retrieve, choose references carefully, avoid asking systems to imitate living creators or reproduce someone else’s work, and be honest about what we can’t trace.”
2. **Attribution: “Who made it?”**    
This is where Jackson argues agencies have more work to do. AI use is increasingly visible. Human contribution often isn’t. “‘Made with ChatGPT’ tells us remarkably little,” she says. “One person accepts the first response. Another brings specialist knowledge, frames the problem, rejects plausible routes and spots the one connection worth pursuing. The missing part of attribution is the interaction: what the person brought, what the model brought, and what happened between them.”

Jackson’s proposed fix isn’t policing AI use so much as evidencing human judgement.

> "For significant AI-mediated work, agencies should preserve more than the final output: enough of the working process to create a contribution record they can audit. Not to assign creativity in percentages, but to make human contribution transparent and inspectable."

That changes the conversation with clients about what, exactly, they’re paying for. “Not: we barely used AI. Not even: look what AI helped us make. But: yes, AI helped make this. And here is what our people actually did.”

Virginie Goupilleau, founder of industry AI consultancy BerylliumIV, sees the same erosion playing out at the channel level. “The old saying ‘half my advertising is wasted; I just don’t know which half’ is about to be a much harder question to answer because of AI,” she says. “Generative engines have become the new intermediary between brand and buyer, and we are still figuring out its working.” When an LLM recommends a brand, the answer draws on brand consistency, PR coverage, reviews, creator content and advocacy, “all blended into one confident answer. Often, there’s no click and no referrer.”

Proving which activity earned that recommendation, she argues, is close to impossible… and the irony cuts deep.

> "The activities hardest to measure (brand, PR, creative) appear to be exactly what the models reward."

Goupilleau’s prescription isn’t a better single metric, but giving up on finding one. “Measurement needs an entire revamp, grounded in systems thinking, so brands can claim effectiveness with confidence. No single method can cut it now, so start orchestrating several. Bring your teams, channels and partners closer together. Then triangulate econometrics for long-term contribution, controlled testing for incrementality, and tracking how often, and how accurately, AI recommends you.” None of it works, she warns, “while teams and channels keep measuring, and claiming credit, in silos.” "

![Virginie Goupilleau](https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/assets/QC0zNJnGM4/virginie-goupilleau-852x958.png)

**Virginie Goupilleau**  
*Founder*  
Industry AI consultancy BerylliumIV

![Amy Banks](https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/assets/kZkElThff8/amy-banks-800x800.jpg)

**Amy Banks**  
*Head of SEO & GEO*  
Havas Media Network UK

Amy Banks, Head of SEO & GEO at Havas Media Network UK, sees the same problem from the sharp end of search. Her advice is practical: “Agencies should prioritise clear authorship and dates, use structured data, publish original research, and keep brand facts consistent across platforms,” she says. “They should also track emerging licensing and attribution models, such as Cloudflare’s ‘Pay per Crawl,’ and help shape industry standards through forums like the IPA.” In an AI-mediated world, she argues, success will favour agencies that can prove what they created, not simply claim credit.

That means being deliberate about what actually gets noticed.

> "Generic content gets absorbed into background information, whereas clear, insightful content stands out."

The assets worth building are “proprietary data, clear structures, relevant quotes, and original research that models can cite or visibly echo.” Her advice is to optimise for being the cited source inside AI answers, rather than simply being present somewhere in the training data, and to shift focus towards demonstrable expertise.

Banks is less optimistic that this will become a solved problem any time soon. Assigning credit, she says, “when AI has an ability to collaborate with original human thinking, direct data input and information ingested from secondary sources on the web, does make it difficult to trace.” Today’s models can’t always reliably trace their own outputs back to source - which is precisely why, in her view, attribution will keep needing governance and regulation rather than a purely technical fix.

**The aerial camera**

![](https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/assets/BE3VXzrDBR/istock-2221641607-1212x866.jpg)

![Marcos Angelides](https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/assets/Zz43Rv6Etk/marcos-angelides-3000x4500.jpg)

**Marcos Angelides**  
*Managing Director*  
L’Oréal Lab,  
*Head of AI Operations* Publicis Media

Marcos Angelides, Managing Director of L’Oréal Lab and Head of AI Operations at Publicis Media, reaches for a different picture entirely: not a stolen joke, but a stolen goal. Picture the final minute of a cup final. The striker rounds the keeper and buries it. The crowd sings one name - but the striker didn’t score alone. The goalkeeper started the move, the defence worked it through midfield, the winger found the final pass. The team created the goal; one player took the shot.

“Marketing measurement often makes the same mistake,” Angelides says. “It celebrates the finish and overlooks the build-up play. Paid search can claim a sale sparked by TV; a performance channel can convert the demand from brand campaigns created months earlier. The touchpoint nearest to the outcome is usually the easiest one to count, even when it’s only the final link in a much longer chain.” AI, he argues, is simply stretching that old weakness further. Recent research from Similarweb found that 91% of AI-influenced traffic is misattributed - 56% to search, 20% to direct. “The influence happened, but our measurement systems may never see it.”

For Angelides, this is as much a provenance problem as a measurement one. “This is partly a measurement problem but it’s also a question of provenance. Where did the information come from, who created it and who receives credit when it’s reused?”

That, he says, puts agencies in a more demanding position than they might realise. “We can’t treat AI-generated answers as evidence simply because it sounds assured. We need to check the original sources, understand whether information has been quoted or transformed, and be clear about the limits of what can be verified. When original thinking, data or creative work has informed an output, its provenance should remain visible. That matters for accuracy, intellectual property and trust.”

Angelides splits the attribution debate into three separate questions often bundled together as one:

1. **Who generated the visit** - classic click attribution, “the part of the problem marketing has spent decades refining.”
2. **Who shaped the decision**, which AI complicates because “a recommendation can happen in one place and the action in another. A shopper may ask an assistant what to buy, then search for a brand, visit a retailer or walk into a shop. The recorded channel captures the action, but not necessarily the influence.”
3. The question Angelides considers most overlooked: **who created the knowledge behind the recommendation in the first place?** “An answer about skincare, market trends or business strategy may draw on publishers, creators, researchers, industry bodies and comparison sites. The response looks singular on screen, but the work behind it is anything but.”

So, should the industry simply measure harder? Partly, he says. “If marketers cannot see where value is being created, they risk moving money towards the channels that receive credit rather than those doing the work. The industry should keep developing new ways to track visibility in AI environments, monitor citations and account for influence that does not arrive through a click. Accountability does not become optional because the customer journey is messy.”

But he’s careful not to let that become the whole answer. “A better attribution model will not turn marketing into a neat sequence of isolated inputs and outputs. The stronger lesson may be that growth has always come from connected systems: media working with creative, data, customer experience and brand building. People do not encounter a spreadsheet of channels. They encounter a brand, repeatedly and in different places, until something prompts them to act.”

His conclusion returns to the pitch.

> "Traditional attribution is like a linesman watching the final pass and the goalscorer. It captures the decisive moment, but most of the move happens before that moment, where the linesman isn’t always watching. What marketers need now is closer to the aerial camera: a view of how the play developed, which players changed direction and where the opening was created."

Done properly, he argues, that wider view “will give every contributor a perfectly calculated share of the goal... AI, search, social, creators, publishers, paid media and brand investment all shape the outcome. The goal still matters. So does the team that made it possible. Because no single player wins the match alone.”

**Or the wrong fight entirely**

![](https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/assets/jX014tFK2J/istock-2215793112-1365x768.jpg)

![Gracie Page-Fozzati](https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/assets/8tCU07Sx73/gracie-2117x3174.png)

**Gracie Page-Fozzati**  
*Senior Global Marketing Director*   
Unbound IA

Not everyone thinks attribution is the fight worth having. Gracie Page-Fozzati, Senior Global Marketing Director at Unbound IA, argues the industry is pouring its energy into the wrong question. “A lot of the industry is pouring energy into attribution: who gets the credit when an AI rebuilds a brand from scraped fragments, half-remembered campaigns and metadata that’s shaky at best and nonexistent at worst,” she says. “I get the instinct, but it’s the wrong fight. Brands still trade value for value, but the currency has changed.”

> “The question isn’t ‘was I credited?’ anymore. It’s ‘was I chosen, recommended, shown the way I’d want to be?’”

Her reasoning draws on how the industry handled the last existential threat to credit and ownership. “Copyright didn’t die with the internet, but it didn’t win on principle either. It got sorted out through licensing, platform deals and Content ID – painful, slow and messy. AI is heading the same way: the people with leverage negotiate, and everyone else gets absorbed. That’s the harsh reality of it. Attribution only pays when you own something the machine actually needs.”

So the brand’s job hasn’t changed, in Page-Fozzati’s view - “show up to the end consumer the way you want to show up, same as it’s always been.” What’s changed is who’s doing the showing. “If the AI everyone’s using rebuilds your brand from its parts, then controlling those parts is the job now. That means clean product facts, language and proof points consistent enough to survive being remixed, third-party signal you’ve earned, and data only you have.”

Brands, she argues, “are going to be more alive and less controllable than ever. That’s ok – if you know how to build for it.” As for agencies: “the job is still to advise, shepherd, steer and deliver clients to growth. Ads have been ripped off since ads were invented. The ones that win are the ones people still recognise.”

**So, who does get the credit?**

![A diagram of the orbits of the planets in the solar system, with a dark background and colorful lines radiating from the center of the circle.](https://the-ipa.shorthandstories.com/the-ipa-ai-magazine-november-2026/future-of-the-industry/the-attribution-gap-who-gets-the-credit-in-an-ai-mediated-world/assets/mPlYnWW9zg/istock-1405085416-1320x795.jpg)

Taken together, there’s no consensus, but there is a pattern. The old, tidy line from idea to outcome is gone. What’s left is a choice: build the evidence trail for the judgement no one can see, or build a brand sturdy enough to survive being pulled apart and reassembled by a machine. The agencies that get ahead of this won’t be the ones with the cleanest attribution model. They’ll be the ones doing both.

## Keep reading

Published by the IPA, the IPA AI Magazine helps member agencies make sense of the fast-changing AI landscape, with practical insights, expert perspectives and real-world applications of generative AI.  
**Managing Editor**: Conor Nichols   
**Design**: Jenna Betts

© 2026 The Institute of Practitioners in Advertising. All rights reserved. No part of this site may be reproduced without our permission.
