Havas Media Network's Amy Banks on SEO, GEO, and Zero Click Search

Havas Media Network's Amy Banks on SEO, GEO, and Zero-Click Search

Amy Banks, Head of SEO and GEO at Havas Media Network, joined ExchangeWire's Lindsay Rowntree to argue that AI-generated answers are absorbing informational queries and that brands need a new practice, Generative Engine Optimization (GEO), to stay visible. Her prescription: restructure content into "chunked" pages that crawlers can pull easily, invest in earned publisher placements, and replace keyword rankings with a share-of-voice metric that tracks how often a brand appears in AI answers.

Worth naming the incentive structure. Banks leads an SEO/GEO practice at a holdco that sells channel expansion, and every recommendation in the conversation expands billable scope. The "lazier crawler" claim is her heuristic, not OpenAI or Anthropic documentation. The claim that affiliate and third-party placements drive citation share ships with no data. None of that is dishonest, but the incentive to name a new discipline you happen to lead is total.

The more interesting thread is the structural one. If AI models cite high-authority domains rather than rank them, editorial authority becomes monetizable again after a decade of programmatic commoditizing it. Publishers who own trusted, frequently-cited domains become retrieval inventory, not just ad inventory. That re-pricing is real regardless of whether GEO becomes a formal agency line item.

The measurement problem is where operators should focus. Havas is building its own citation-share tooling precisely because no neutral third-party product exists. An agency grading its own GEO work is a conflict buyers have rejected in every prior search cycle, from keyword rank trackers to viewability verification. The pattern is consistent: a new surface appears, agencies improvise measurement, then a neutral vendor productizes it and sells to both sides. Our call: by Q3 2027, at least one of DoubleVerify (brand-safety and viewability measurement), Integral Ad Science, or Comscore will launch or acquire a named AI-citation visibility product. All three already have the crawling and reporting infrastructure to add citation tracking cheaply, and leaving the category to holdco in-house tools is not how any of them operate. Medium confidence, because timing could slip.

For operators today, the spend calculus is simple. There is no paid GEO channel at scale yet, so the real cost is opportunity cost: staff hours pulled from paid search that still converts. The one low-regret move is content hygiene on your own domain, because it helps whether or not GEO becomes a real budget line. Do that. Watch earned publisher placements as the one channel genuinely gaining value. And refuse to buy anyone's citation-share dashboard until a vendor without a stake in the outcome is doing the measuring.

Amy Banks, Head of SEO and GEO at Havas Media Network, told ExchangeWire's Lindsay Rowntree that AI answers are eating the informational query, and her fix is to treat Generative Engine Optimization as an add-on to the SEO team, retooled around "chunked" content, earned publisher placements, and a share-of-voice metric that replaces keyword rankings. The question for an operator: is GEO a real budget line forming, or an agency repackaging organic search to keep a shrinking practice billable?

What's actually being decided: whether publishers, agencies, and measurement vendors should invest now in optimizing for and measuring AI-answer visibility, or wait until the money and the standards show up. Type 2, reversible. Nobody is betting the company on GEO in 2026. The cost of waiting a quarter is low. The cost of building the wrong measurement stack is higher.

Forcing function: none, really. There is no upfront, no renewal cliff, no regulatory date driving this. Which shapes how much you should spend today.

---

The Market Analyst. Follow who wins if zero-click is real. The loser is paid search click volume, and by extension every DSP and search desk that resells Google intent. The winner is anyone who owns a high-authority domain the models cite. That's a quiet re-rating of premium publishers as training and retrieval inventory, not just ad inventory. Banks flagging affiliate and third-party placements as the channel with "much bigger impact" is the interesting thread here. If earned publisher mentions drive citation share, editorial authority becomes a monetizable asset again, after a decade of programmatic commoditizing it. For a generalist reader: the websites the AI trusts get quoted, and being quoted is the new being ranked.

The Skeptic. Steelman the case against GEO as a category. Banks runs an SEO/GEO practice at a holdco that sells channel expansion. Every claim in this episode points toward more billable scope: chunk your content, buy earned placements, license our proprietary share-of-voice tool. The "lazier crawler" claim is her own heuristic, not OpenAI or Anthropic documentation. The affiliate-lift claim ships with zero data. None of this is dishonest, but the incentive to name a new discipline you happen to lead is total. What has to be true for GEO to matter? That AI answers meaningfully move purchase behavior, and that a brand can influence what the model says. Neither is proven in this conversation.

The Operator. Tuesday morning, what breaks? You reallocate content spend toward "chunked" pages and earned placements, and 90 days later you cannot prove any of it worked, because the only measurement is Havas's own tool grading Havas's own work. That's the seller scoring the exam. Second-order effect: your SEO team, already stretched, now owns a second discipline with no standard tooling and a moving target, because the platforms change retrieval behavior with no changelog. The one concrete, low-regret move is content hygiene on your own domain. That helps whether or not GEO becomes a real line item, so do it regardless.

The Customer / End User. Two customers here. The brand advertiser wants to know if being cited by ChatGPT sells anything, and nobody in this episode answers that. The consumer just wants the answer without ten blue links, and increasingly gets it. That consumer behavior is the only thing in this story not talking its book. When people stop clicking through for "what's the best X," the publisher loses the pageview and the affiliate loses the referral, even if the citation stays. Citation share without a click is a vanity metric until someone ties it to a conversion.

The CFO. Show me payback. ChatGPT paid ads are, per Rowntree's own demo, "quite primitive." So there is no paid GEO channel to fund at scale yet, only organic effort and a measurement license. The real cost is opportunity cost: staff hours pulled from paid search that still converts today, spent on a channel that pays back on an unknown schedule. The rational spend is small and instrumented. Fund the content hygiene, buy or build a cheap tracking read, and refuse to write a big check until citation share moves a sale.

---

Where the council splits. The Market Analyst sees a structural shift worth positioning for early: premium domains becoming AI-retrieval inventory. The Skeptic and the CFO see an agency expanding scope ahead of any proven ROI. That's the real disagreement, and it hinges on one fact nobody has: does an AI citation move a purchase? The second tension is measurement. The Operator won't trust a vendor grading its own work; the Market Analyst sees exactly that gap as the white space a real third-party measurement player should fill.

What it hinges on. Two beliefs. One, that zero-click keeps absorbing high-intent queries and not just idle informational ones. Two, that brands can reliably influence model output and tie it to conversion. The council leans cautious-but-moving: do the free stuff now (own-domain content hygiene, a rough citation read), refuse the expensive stuff (big paid-GEO budgets, sole-source measurement) until an independent metric exists. The genuinely investable read is the Analyst's: earned publisher placements getting a second life. That one has legs beyond the agency pitch, because it re-prices an asset publishers already own.

The measurement vacuum is where the real opportunity sits. Every prior search era got a neutral referee: keyword rank trackers, then verification vendors. AI-answer visibility has none, and agencies are filling it with in-house tools that grade their own work. That never holds. Buyers eventually demand a third party.

Prediction: By the end of the 2027 upfront season (roughly Q3 2027), at least one established measurement or verification vendor among DoubleVerify, Integral Ad Science, or Comscore will launch or acquire a branded AI-answer / LLM citation-visibility product, positioning it as the neutral alternative to agency-built tools like Havas's.

Confidence: Medium. The vacuum is real and these vendors chase exactly this white space, but timing could slip past the upfront.

Why: Havas is building proprietary citation-share tooling precisely because no standard third-party product exists yet, and an agency measuring its own GEO work is a conflict buyers have rejected in every prior search era. The verification vendors live off exactly this pattern: a new surface appears, agencies improvise measurement, then a neutral referee productizes it and sells to both sides. DoubleVerify, IAS, and Comscore already sell brand-safety and visibility measurement and have the crawling and reporting infrastructure to bolt on citation tracking cheaply. The less likely outcome is that they sit still, because leaving a measurement category to holdco in-house tools cedes ground they normally rush to claim.

Revisit by 2027-09-30: We're right if DoubleVerify, IAS, or Comscore ships or buys a named AI-citation / LLM-visibility measurement product by then. We're wrong if the only such tools in market by that date remain agency-built or startup-only, with none of the three moving.

Net for the operator: impact today is low, and the episode is more orientation than signal. Do the free work on your own domain, watch earned publisher placements as the one channel genuinely gaining value, and don't buy anybody's citation-share dashboard until it isn't the seller grading itself.