Let's Talk

THE SIGNAL - JULY 2026 AI

AI AMA with Wil Reynolds and Alisa Scharf

Thursday, July 16th @ 12pm ET

This month Wil and Alisa dug into the early data on ChatGPT ads, how branded query fan-outs are hardwiring certain brands to certain topics, and why a visibility report is a trap if you can't measure GEO against the business.

Plus: what to do about all three.

Missed the live session? Join us next month.

Get Transcript View Slides

Play July Signal Audio

57:50

July 2026 Insights and Action Items

1. ChatGPT ads are here, and the early data says this is more than an awareness play.
What's happening: Our team has been piloting ChatGPT ads for clients, and with blended click-through rates running higher than expected and CPM and CPC data now showing up as of June, the biggest early lesson is to bid to the prompt, not the page, since your ad lands at the bottom of a live conversation.
Action: Run the prompts that matter to your brand in a free ChatGPT account and see whose ad shows up.
2. A query fan-out is the model's behind-the-scenes research, and brands are now showing up inside it.
What's happening: When an LLM answers it runs its own set of searches to validate the response, and we ran one prompt 30 times to find iPullRank in the fan-outs 97% of the time and Mike King 63%, a sign that certain brands are getting hardwired to certain topics through years of publishing a real point of view.
Action: Pick 3 priority prompts, run them through SE Ranking’s query fan out tool, assess ChatGPT’s association of the prompts with your brand (and your competitors’ brands).
3. Visibility reports are a trap if you don't tie them to an outcome.
What's happening: Measuring GEO is hard, and if you hand an executive a visibility report they'll love it for three or four months and then ask what it did for the business, so overlay visibility against a real outcome like traffic, signups, or conversions to catch when visibility climbs but the business impact doesn't follow.
Action: What metric do you wish you could measure? Come back next month with your idea for new measurement KPIs - starting with the next metric.

Q: Is it worth chasing citations and mentions for non-branded, informational AI prompts? A: Only once you've tightened up your brand, which isn't what people want to hear. Start by checking what AI says about you versus your top competitor. If the branding already worked, people know both brands, so make sure the models represent yours accurately before wading into the unbranded quagmire of shifting models, fan-outs, and entrenched players. When you do go after non-branded prompts, be brave enough to focus on the things you and your client both know you genuinely crush, confirmed by customers and not the marketing team. If you can't win where you're strongest, you won't win where you're weak, and the models will eventually catch that anyway.

Q: What baseline KPIs resonate with leaders who are new to GEO but want a realistic view? A: The one Alisa likes is branded search, both paid and organic, tracked over time. The hypothesis: when someone discovers a brand on ChatGPT, they often run a branded search to validate it, so branded impressions and clicks should climb as your AI visibility does. It's a moving target that favors marketers right now, since more of your audience adopts AI search every month. Just account for other variables, like a bump in branded spend or a new awareness campaign.

Q: What is GEO doing to the rest of the funnel, now that buyers seem pre-sold from the ChatGPT conversation? A: It's flattening the funnel in some ways and just reordering it in others. As Marketa has said, the funnel hasn't looked like our illustrations in quite a while. A search that used to start on Google now starts on ChatGPT, and that person may still circle back to Google to validate what they learned, or may not come back at all.

Q: How do you pick the right prompt list, by intent, category, region, branded vs. non-branded? A: All of the above, approached both diagnostically and by hypothesis. Add persona to the list, benchmark the brand's current visibility, and track the specific tests you're running so you can confirm or deny each hypothesis. Check out more thoughts on this from the Seer team.

Q: Couldn't you just benchmark traffic, track your GEO mentions, and watch whether sales rise? A: You can, but you hit the same attribution trap as SEO. Fix canonicals, update metadata, and earn a few links in the same window and you can't say which one moved traffic. It only works cleanly in a vacuum where GEO is the only thing changing, which never survives C-suite scrutiny once they factor in partnerships, tech changes, or site speed. In retail especially, with year-over-year and quarterly goals, the real fix is educating leadership that the environment changed, then supplementing the old measures with new ones instead of yanking the rug out.

Q: Some well-established, highly capable companies still aren't recommended by LLMs. Positioning, entity understanding, or technical issue? A: Alisa's first check is technical: did the brand block ChatGPT or Claude bots early on? Long-tenured sites sometimes did. IMDB is the classic case, built for crawlers and essential to the web, where Google found a workaround but AI search hasn't. So start by asking whether something is stopping the models from understanding the brand at all.

Q: Do you talk to clients about how they show up in the training data? A: Not really, because none of us have a clear read on it. If you had to, look at generated responses with no web citations and no live search, and note which brands appear on branded queries, including competitors'. Be honest that it's close to a fool's errand: a new model can drop next week with refreshed data nobody announces, and Gemini, ChatGPT, and Claude all pull from different training sets. Fine for a one-off to win stakeholder buy-in, not worth chasing.

Q: Does the reasoning in these models come from the same people who built the pre-training knowledge? A: No. These models are built differently from search engines, both in what they come preloaded with and in how they generate responses. Frontier models invest heavily in post-training, using human experts to define what good and bad answers look like, but those aren't the same people behind the pre-training corpus. Many hidden layers combine to produce an answer, and trying to detect all of it does marketers a disservice. We highly recommend this video from Andrej Karpathy on Intro to LLMs.

Q: Are Bing's AI impressions really brand mentions, or just citations? A: Citations. The Bing data is measuring citations.

Q: How important is it to use long-tail keywords verbatim to hit the projected ranking outcomes? A: Increasingly less important. These models deeply understand the semantic connection between different ways of phrasing the same thing, so lean toward how your audience actually refers to something, and at minimum include both their phrasing and yours. Worth testing if you stray far, but the need for verbatim keywords is fading.

Q: What qualifies as LLM-friendly content? Beyond listicles, how do you communicate good info for AI tools? A: Whatever your audience is looking for that isn't already answered. Find it by sleuthing Reddit and forums, and by mining your call center and support tickets for the questions people actually ask. If you have none of that, watch where your community gathers and interview your audience directly. The more you publish, the more reactions you get to sharpen the next round.

Q: Is it still worth keeping a bunch of regular posts, especially AI-generated ones maybe read only by bots? A: As framed, it doesn't sound valuable, but it depends on what "a bunch of regular posts" actually means. Wil recommends using this audit process to identify content that people actually want to share.

Q: What about small businesses that are highly niche or localized but low on resources? How do you do this work? A: Alisa reframes fewer resources as better constraints: if you're the only marketer and the directly responsible individual, you skip the bureaucracy and can move fast, so lead with what's genuinely unique and be risk-tolerant about sharing your proprietary insights. Wil's close: take ownership like a DRI ("somebody bought your service this week, they just didn't buy it from you"), and know that AI now rewards receipts. Big brands used to slap thin claims on a page and outrank honest operators; now the models look for proof, which gives a nimble, genuinely good small business a real shot. Be honest about where you're not the best, then find the niche that loves what you do and speak to them.

Wil and Alisa's Latest Thoughts

How an intern 10x'ed impressions with 1 AI experiment

TL;DR: Bryan Gunawan, an intern in Seer's AI Optimization Academy, made 5 targeted changes to an existing Seer blog post and took it from under 100 impressions …

Content Strategy / People Of Seer / Seo Strategy / Geo

Read more →

My GEO Strategy & Consulting Tool is open sourced

This GEO consulting tool is built for strategists who struggle, like me, to articulate what tactics we have to deploy in GEO against what expectations should be…

Geo

Read more →

Can a prompt undo years brand building? Our AI interview tests say yes

TL;DR: We interviewed 28 people across almost 100 tasks, having them use AI to solve a product selection problem, and we ask them what brands people intend to b…

Ai / Geo

Read more →

The Proof: ChatGPT 5.5's fanout patterns reveal the importance of brand

If you've been looking at the whole "do I have to do listicles" thing and throwing up in your mouth a little, this post is for you.

Ai / Geo

Read more →

About the Speakers

Wil Reynolds

CEO, Seer Interactive

Wil has over 15 years of experience doing and talking about digital marketing. He loves to speak at conferences about the future of search. He also enjoys sharing his ideas and marketing insights with his 42K LinkedIn followers.

Alisa Scharf

Chief AI & Innovation Officer

Alisa Scharf has with over a decade helping Seer clients and teams navigate the biggest shifts in digital marketing. She's a regular conference speaker on AI-driven search and the value of operationalizing AI within organizations.

Get insights like this in your inbox

AI-era marketing strategy, tools, and tactics. No fluff.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

The Signal - July 2026

July 2026 Insights and Action Items 1. ChatGPT ads are here, and the early data says this is more than an awareness play. What's happening: Our team has been pi…

Events & Webinars / On-demand

Read more →

On-Demand: The Signal - June 2026

Seer's AI "Need-To-Knows" for June 2026 1. Bing Webmaster Tools now gives you grounding queries. It's the first to do it. What's happening: ChatGPT, Claude, Gem…

Events & Webinars / On-demand

Read more →

On-Demand: The Signal - May 2026

Events & Webinars / On-demand

Read more →