Insights

Re-imagining AI Visibility KPIs in a Query Fan Out world

Admin — Fri, 02 Oct 2026 16:30:48 GMT

Wil, take me right to the KPI for long term sustainable GEO visibility!!

SaaS Marketers - Promote your Claude & Codex connectors & win in AI search

Admin — Mon, 28 Sep 2026 20:30:29 GMT

Claude and I had a 12,000 word back and forth on how to use Cloudflare and GA4 data to answer a hard question. If you were a SaaS marketer what would you do with all that context to make better answers?

AI Citations Aren't Your Top KPI: ChatGPT-5.6 Luna Runs More Targeted Searches

Admin — Thu, 24 Sep 2026 19:36:11 GMT

TL;DR

We ran a direct comparison of 346 matched prompts to see whether ChatGPT-5.6 Luna behaved differently from GPT-5.5
GPT-5.6 Luna used web search more often but ran fewer queries, resulting in little movement on the number of fan-out queries compared to GPT-5.5
Domain-target searching increased drastically (more than 3x), while consulted sources and inline citations dropped
Our take: Luna is getting better at finding information and needs fewer resources, so brands should pay more attention to the full retrieval process to measure AI search visibility

How Do ChatGPT-5.5 and GPT-5.6 Differ?

We’ve researched ChatGPT-5.5 extensively, including a deep dive into fan-out patterns. Now that GPT-5.6 Luna is the default model as of July 9, 2025, it's the model people get when interacting with ChatGPT and the model brands will be reporting on through AI visibility trackers.

With that in mind, we wanted to know how the two models compared. Specifically:

How did the retrieval process change across domains and citations?
Does GPT-5.6 Luna search similarly, or were there notable differences?
Did fan-out query behavior shift or stay the same?

We compared GPT-5.5 and GPT-5.6 Luna across 346 matched prompts from six industries.

Spoiler alert: At first glance, the two models looked very similar. Across the 346 matched prompts, they ran almost exactly the same number of fan-out queries. The difference is how targeted Luna's searches are. It used the site: operator in 76% of its fan-out queries, up from 22% with GPT-5.5.

That leaves far fewer open searches for everyone else. So if Luna isn't already searching your site directly, your barrier to showing up is higher than it used to be. Check out our top findings below.

1. GPT-5.6 Luna Uses Web Search More Often, But Runs Fewer Queries Per Search

GPT-5.6 Luna turned to web search more often than GPT-5.5. Across the six industries, Luna searched on 85.2% of calls, up from 80.8%. But each time it searched, it ran fewer fan-out queries: 3.42 per search, down from 3.63.

Those two changes almost exactly cancel each other out. Luna searched more often but ran fewer queries each time, so the total number of fan-out queries barely moved. GPT-5.5 ran 733 and Luna ran 729, a difference of just four.

We thought there would be more fan-out queries after the model update, especially given the increase we saw when GPT-5.5 was released. Instead, the total stayed the same, and the change showed up in how Luna searched.

Our hypothesis is that Luna needs fewer queries per search because its queries are more targeted. When a fan-out query goes straight to a specific website (more on that in the next finding), one query can do the job that used to take several broader ones. So the two changes balance out in volume, but not in effect. More prompts now send ChatGPT to the web, and each search covers less ground.

2. Domain-Targeted Searching More Than Tripled, Especially on Non-Brand Prompts

GPT-5.5 had already begun using the site: operator (a search command that limits results to a single website) in fan-out queries to target specific brand sites for answers and, in our opinion, prevent low-quality responses. GPT-5.6 Luna elevated this retrieval method even further.

GPT-5.5 conducted site: searches in about 22% of its fan-out queries, while Luna did so in 76%.

On branded prompts, third-party sites (publishers, review sites, research orgs, etc.) appeared at roughly the same rate with both models: 26.6% with GPT-5.5 and 24.4% with Luna. So even when a user asks about your brand directly, ChatGPT still looks beyond your website.

That means the perception of your brand outside of your website matters just as much as the narrative you share on sources you control.

The actual jump came from prompts that didn't mention a brand. Non-brand prompts saw a rise in site: searches, from 20.6% in GPT-5.5 to 77.3% in GPT-5.6. That means Luna was the one to decide on a domain to search even when the user didn’t specify a brand or website.

Since ChatGPT is getting more targeted with its retrieval, the overall number of consulted sources (the sources ChatGPT pulls in while searching, before it decides which ones to cite) fell 43%, from 6,134 with GPT-5.5 to 3,481 with Luna.

This presents a unique opportunity for brands launching AI search programs. If GPT-5.6 Luna is already targeting your site with site: searches, make sure the information it's looking for is easy for both people and AI to find.

If your brand isn't one of the sites Luna searches directly, you may see fewer citations now that it's the default model. Open searches, the ones any site could appear in, fell from 575 to 175, so getting results will likely take more work.

Look at every stage of retrieval, from fan-out queries to consulted sources to citations, to see how and where ChatGPT finds information about your brand.

3. Inline Citations Fell 13%

We know Luna searched in a more targeted way and consulted fewer sources. In the final answers, it also gave fewer inline citations (the sources linked in the answer itself).

Across all 346 matched prompts, GPT-5.5 produced 2,054 inline citations, while Luna produced 1,790, a decrease of 12.9%. Searching more often didn't translate into more citations.

Our take is that Luna has gotten better at knowing where to find information, so it needs fewer sources to back up its answers.

This matters for any AI search program that uses citations as a KPI or as the success metric for an experiment. If your citations dropped after Luna became the default, that may have nothing to do with your strategy or the tests you have running.

Think of it like a Google algorithm update: the platform changed, and performance shifted with it. Mark the switch to Luna in your reporting so a platform-wide drop doesn't get read as a failed experiment.

So What Should You Do Now?

How can you apply these findings to your AI search strategy? Here are my recommendations:

Look further down the retrieval process when you’re measuring AI search visibility. Most AI visibility trackers only report citations, the bottom line of the retrieval process.

The bigger benefit comes from the layers behind them: the fan-out queries ChatGPT runs, the domains it targets with site: searches, and the sources it consults before deciding what to cite. That's where you'll see why your citations moved, not just that they did.

Pay closer attention to the connections the models appear to make between brands and topics. Luna often decides which sites to search before it retrieves anything, so being recognized as a relevant source for a topic matters before any page gets retrieved.

Run a base set of nonbrand prompts you’ve been monitoring to capture the fanout queries and sort the site: searches by topic. Which topics does ChatGPT search your domain for, and which does it search your competitors' domains for instead?

Compare across models or over time, and you'll see where you're gaining ground and where a competitor owns the association. The topics your competitors own are where you should focus next: both in your own content, and in your coverage on publishers, review sites, and research organizations.

Ready to level up your AI search strategy but don’t know where to start? The Seer team has extensively researched what works and what’s a short-term trend, and we can recommend where to put your resources based on your industry, budget, and team size. Chat with us to learn more.

Our Most Valuable Blogs of 2026: According to Our Readers

Hannah Cooley — Tue, 22 Sep 2026 21:00:18 GMT

What's my favorite horror movie? When I patted myself on the back for setting up an ~autonomous workflow~ that kept our newsletter subscriber hygiene clean by automatically unsubscribing anyone who hasn't opened any of our emails in the past 12 months.

Are You Struggling with Market Pressure or Operational Readiness? Use Our 2x2 Framework to Find Out

Admin — Thu, 10 Sep 2026 16:03:54 GMT

TL;DR

If your performance numbers are down, your single-axis maturity model might be telling you to solve for the wrong problem
A true framework should factor in two variables: what the market is doing to you (market pressure), and what you're equipped to do about it (operational readiness)
Use our 2x2 framework (and our handy worksheet) that considers both variables to more accurately evaluate where you fit
Two Seer clients thought they had performance problems but used our framework to pinpoint other underlying issues
Learn three quick things you can do this week to help identify the underlying issues that could be impacting your numbers

Is Your Maturity Score Focusing on the Wrong Problem?

I've worked with dozens of marketing teams, and every one of them has had the same bad week. Numbers dip, costs climb, and someone asks what happened.

Google Is Not Going to Stop Ranking Your Claude-Assisted Content

Admin — Fri, 04 Sep 2026 18:24:31 GMT

TL;DR

Claude’s watermark only flags that Claude was “likely involved,” so it can’t tell a fully AI-drafted piece from a human draft Claude lightly edited
It’s an EU compliance requirement (Code of Practice on Transparency of AI-Generated Content). This is not a new Google ranking signal.
Google has run its own version (SynthID-text)on Gemini for two years and never used it to rank anything
The real traffic risk is thin, templated content and not the watermark. Sites that scaled content off a copy-the-template approach saw more than half those lose 30%+ of peak organic traffic
What still matters: unique inputs and a human accountable for the final piece

Anthropic started watermarking Claude's text output.

AI Can Do More for Project Management Than Note-Taking: 4 Ways to Level-Up Your Usage

Admin — Wed, 02 Sep 2026 18:46:20 GMT

TL;DR

AI is more than just a summary or synthesis tool for project managers, and the best PMs will use it to become a better strategic partner
PMs can use AI to spot ongoing friction and pinpoint the root cause, or look for and resolve broader patterns across your client accounts
Use AI as a sounding board to find the gap in your idea before bringing it to the client
Leverage internal expertise and documentation to build client-ready deliverables with minimal manpower
AI can't interpret human behavior or responses, and that's where PMs can shine when they offload busywork to AI

Are PMs Extracting the Maximum Value from AI?

By now, most project managers are using AI to pull action items from a call, draft a status update, or help you write an SOP.

But if the tactical execution part of the job has gotten cheaper and faster thanks to AI, what's left for a PM to be good at?

I think the answer is the thing most of us got into this work hoping to do in the first place: being a genuine strategic partner to the client, not just the person who keeps the trains running on time.

Here are 4 ways I’ve started using AI to level-up my client strategies and build stronger client relationships.

1. Catching the Pattern You'd Never Notice Call by Call

Our CEO, Wil Reynolds, says something that’s stuck with me: use AI for the manual tasks your brain power far exceeds, and spend that brain power on strategy and moving the needle for the client. I didn’t fully get what that meant until I started actually doing it.

Now it's part of my process. After almost every client call, I ask AI one question based on the transcript:

THE PROMPT
How can we be a better strategic partner here, based on what the client said?

Just asking AI to dig for subtext instead of a summary isn't a differentiator anymore. Plenty of PMs have figured out that part. What's more impactful is what you do with whatever AI surfaces.

In practice: With that prompt, AI helped me understand the root of recurring campaign delays for one client: they didn’t know where our campaigns stood, and approval requests were scattered across Slack threads with no clear queue. They weren’t stalling on purpose, they just had no single place to understand what was needed from them.

After AI picked up on the pattern, I worked on a solution and built a live campaign tracker. Now, the client had:

A scheduled weekly update via Slack on where every campaign stood
A dedicated approvals-only Slack channel for our team to share exactly what we needed from them
Updates formatted so the client could scan, review, and approve in minutes

The client stopped being an accidental bottleneck and started closing out approvals in minutes, and our team stopped losing time chasing sign-off across multiple channels.

A single call would never have told me that on its own. It took AI flagging the same friction call after call before the real fix became obvious.

TRY THIS
After your next few client calls, run each transcript through the same question. Focus not on what was said, but on how you could be a better strategic partner. Don't act on the first answer. Watch for the same friction showing up more than once before you build anything around it.

2. Seeing Across Accounts, Not Just Within One

Most PMs manage more than one relationship, and the thing AI is uniquely good at that a busy human isn't, is holding all of them in view at once.

So I started asking a version of the same question I use on individual calls, but pointed at my whole book of business: is there anything showing up across more than one of my accounts that I'm not connecting?

In practice: AI surfaced that GEO test implementation was stalling for a similar reason across a few accounts that, on paper, had nothing in common. It wasn't a technical blocker. It was resourcing; client teams didn't have the bandwidth to execute anything complex, and that was only visible once I stopped looking at each account in isolation.

Seeing that bigger issue led me to change how I approached a GEO project for one client in particular. Implementation had already stalled. Tests weren't moving because the work assumed more bandwidth than the client's team actually had.

Instead of continuing to push on the same complex tests, I raised the resourcing question directly. The client confirmed their team's bandwidth was limited right now, which told me exactly what needed to change: I had my internal team shift focus to easy, low-lift test ideas the client could realistically execute. That shift is what got the project back on track.

That pattern only showed up because AI helped me look beyond my individual accounts when I didn’t have the bandwidth to hold everything in view all at once.

TRY THIS
If you manage more than one account, ask AI a version of this before your next resourcing or roadmap conversation: is anything showing up across more than one of my accounts that I've been treating as separate? Do it before something stalls, not after.

3. Using AI to Pressure-Test Yourself Before the Client Does

AI has a default instinct to agree with you. It wants to be helpful, which in practice often means it wants to validate whatever you just said. Left unchecked, that's the opposite of useful when you're trying to stress-test a strategy.

So I set a standing rule for how mine should behave:

Don't agree by default
Find the weakest point in what I've said before you affirm any of it
Tell me what someone who disagrees would say

When I bring AI a strategy or ask for a read on a situation, its first move is to argue with me, not validate me. That changes the conversation completely. Instead of a tool that confirms what I already think, it's one that makes me defend my idea first.

In practice: A new contact on one of my accounts wanted the relationship with our team to feel more collaborative. She was interested in less status reporting and more working through things together.

My first instinct was straightforward: change our current meeting format. Turn the recurring status call into more of a strategic discussion.

I ran that call's transcript through AI afterward, and it didn't just validate the plan; it found the gap in it. Buried in the same conversation, my own team had voiced frustration about how campaigns were getting launched: almost no input from us on strategy, and asset timelines so tight there was no room to do more than execute whatever landed.

AI's read: changing the meeting format was a reasonable start, but it wouldn't touch the actual problem. If we wanted to be real collaborators, the meetings needed space to shape campaign strategy together, talk through how assets should be positioned by platform, and get ahead of sizing and formatting before assets showed up with days to spare.

That's the version we actually built: a new SOP centered on recurring strategy sessions, not a lighter-touch status call. The relationship is strong enough now that the client's asking about expanding our scope to additional divisions because we became the partner she wanted.

My instinct wasn't wrong. But the fix needed to go further, and I wouldn't have caught that on a single read of my own notes.

TRY THIS
Before you bring your next recommendation to a client, run it past AI with the same standing instruction: don't agree, find the weakest point, tell me what someone who disagrees would say. If it doesn't push back at all, that's worth noticing as it usually means the instruction didn't take, not that your plan is airtight.

4. Spearheading the Work Instead of Coordinating It

The clearest example of what this actually changes for a PM happened when a client asked for a full deck to present in a CMO meeting that had landed on their calendar with about 24 hours' notice. My team was stretched thin elsewhere, and we didn’t have the bandwidth to pull together a normal cross-functional build.

In practice: I built the deck myself by taking our existing analysis decks, call transcripts, and Slack threads for the account, feeding all of it into AI, and specifying the story I wanted the deck to tell. Not “summarize this,” but the actual narrative: here’s what matters to this CMO, here’s the through-line I want a stakeholder walking away with.

The point wasn’t about AI being fast. It was that I had the full context on the account: the relationship history, the priorities, and the political read on the room. AI closed the gap between having that context and having a polished deliverable, without needing to pull in three other people to get there.

It used to take a full cross-functional build to turn scattered account knowledge into a client-ready strategic narrative. Increasingly, a PM who already holds that knowledge can spearhead it, carrying the team’s work into the room rather than convening the room to package it.
To be clear: AI wasn’t replacing our specialists, but rather leveraging the work our specialists had already done to inform the output.

Those analysis decks, the account knowledge, the read on the CMO all came from the team’s expertise. Because I already had the information, I was able to turn it into a polished, client-ready narrative on a 24-hour turnaround.

The layer this shrinks is synthesis and assembly, not subject-matter expertise. And that’s the opportunity: a lot of cross-functional time gets spent packaging what someone already knows. If the PM can spearhead that layer, our specialists get more of their time back for the deep work only they can do.

TRY THIS
Next time you’re under deadline with everything you need already scattered across decks, transcripts, and Slack, don’t ask AI to summarize it. Tell it the actual narrative you want a stakeholder to walk away with, and let it build toward that. See how far you get solo before asking for backup.

What AI Still Can’t Do: Interpret Human Behavior or Responses

None of this makes the PM optional.

For example, go back to that campaign tracker and approvals channel from Pattern 1: AI could see that requests were stuck and tell me exactly where. But it couldn't tell me whether raising that with the client directly would land as helpful or harmful.

That read came from a dozen calls with the client, understanding: how he takes feedback, what the relationship could hold at that point, whether this was a conversation for a call or a quiet Slack message. None of that lives anywhere AI can find it.

AI is good at telling you something's there. It has no way of telling you how it'll land with this specific person, right now. That part is still entirely yours.

Where This Leaves the Project Manager Role

For years, the job rewarded whoever could juggle the most balls. That was never the point.

The PMs who get the most out of AI are using the space that creates to actually think: about the account, about the client's business, about what they're not being told directly.

Here's where I think this actually takes the role over the next year:

Less time spent proving the work got done
Reading what a client isn't saying
Knowing when a relationship needs a different kind of attention than a status update
Building the judgment calls that no prompt can make for you

The PMs building habits like the ones above will spend more time doing the work that made them want this job in the first place, while the busywork keeps quietly shrinking around them.

That's the opening in front of every PM right now: less of your day spent holding it together, more of it spent actually thinking, and finally the room to become the strategic partner you always wanted to be.

If you're looking for a strategic partner to level-up your marketing activities, Seer can help. Chat with our team to learn what we can achieve together.

AI Visibility Is Lying to You: There’s No One-Size-Fits-All AI Search KPI

Admin — Mon, 31 Aug 2026 19:59:48 GMT

TL;DR

Everyone measuring AI search is defaulting to the same KPIs, but there is no one-size-fits-all AI search KPI
New research shows 54% of AI search categories lack a clear category owner, and those categories represent 89% of all AI search demand
If someone else owns your category, you should stop chasing visibility and instead focus on accuracy: when you do show up, does the model describe you the way you actually want to be described?
Accurately measuring AI search KPIs requires a brand canon, a living document of your brand’s claims, positioning, facts, and more

Your Visibility Score Can Be Perfect and Still Be Wrong

AI can mention a brand in every prompt you track, but that brand can still be losing.

Study: AI Declined to Answer Prompts About Our Brand 31% of the Time (But Got Our Info Mostly Right When It Answered)

Admin — Thu, 27 Aug 2026 18:59:59 GMT

TL;DR

We realized AI platforms were relaying an inaccurate narrative about Seer Interactive, so we ran a full analysis to find out what else they were saying — and whether the info was accurate
Of the 6 AI platforms we tracked, they averaged a 95.9% accuracy rating when they answered the prompt
The AI platforms were more likely to decline to answer a prompt (31% on average) than to provide an incorrect answer about Seer (3%)
AI frequently declined to answer prompts about comparisons (80%), but had high accuracy rates for both direct (90%) and indirect (89%) prompts
We share a framework to help brands prioritize inaccuracies in need of fixing, based on how important a claim is, whether it's an easy fix, and if it's objective or subjective

Do You Know What AI Is Saying About Your Brand?

A few months ago, we asked AI models, “what do people say about working with Seer Interactive?” Most mentioned high account manager turnover, which was a surprise to us.

How we built an agentic loop to improve every skill in our company, instantly

Admin — Thu, 20 Aug 2026 18:48:49 GMT

TL;DR

I reviewed our traffic-decline-analysis skill's and found four real problems: a traffic number from the wrong source, a real drop mislabeled "stable," no way to trace where a number came from, and no long-term trend view
I gave that feedback in a 9-minute recording and an agent picked up my feedback and fixed it the same day, and it shipped to everyone at Seer.
I found another problem the skill was picking which checks to run instead of running all of them, every time
An offhand idea from that same review became a new feature days later. The tool now runs off a simple spreadsheet export instead of needing live client data

Here’s how it all happened…

What Is AI Saying About Your Brand? Develop a Brand Canon and Find Out.

Admin — Mon, 17 Aug 2026 16:37:16 GMT

TL;DR

Brands are measuring AI visibility without tracking if the perception of their brand is even accurate
We recommend building a brand canon, which tracks facts, differentiators, brand claims, what your brand wants to be known for, and what you don’t
Once you’ve created your canon, you can use it to compare what you want AI to say about your brand versus what it actually says (and fix the output by tracing claims to their source)
Learn how to build your own brand canon, how to measure the accuracy of claims made by AI, and how to prioritize fixing falsehoods and inaccuracies

What Happens if AI Shares Inaccurate Info About Your Brand?

A potential customer asks AI what your company does, and the answer comes back about 70% right (per Seer's analysis of 28,123 responses about our brand across 6 AI models). But nearly 30% of the information is wrong or outdated, or the model declines to answer.

27 GEO RFP questions you should be asking in 2026

Admin — Mon, 10 Aug 2026 18:37:44 GMT

We looked back at hundreds of sales calls about GEO and looked for the questions people are asking, the pressures they are under, and decided to write up the questions people are asking and the questions they should be asking in an RFP situation.

Like my SEO RFP guide, this guide isn’t limited to those of you looking for an RFP template for your upcoming GEO/AEO project:

Do LLMs Respect robots.txt? Where robots.txt Can and Can't Block AI

Admin — Fri, 07 Aug 2026 18:49:49 GMT

You’ve probably noindex-ed pages you don’t want showing up in search. It’s a solid practice for product variants or paid destination pages, and it works for Google rankings.

Our Early View of ChatGPT Ads: 5 Initial Observations - Updated July 28, 2026

Admin — Tue, 28 Jul 2026 04:00:00 GMT

When we first published this on July 15, we were working from three days of ChatGPT ads data. Revisiting the data this week, our main takeaways haven't changed, but ecosystem is evolving. Share of responses with ads rose significantly, a wider set of brands is now buying them, and early patterns are sharpening.

The updated numbers below now reflect about 2 weeks of data. With that, we advise you treat this as a developing study. It's still early and we are prepared for this to change fast.

OpenAI officially launched the ChatGPT Ads pilot on February 9, 2026, rolling the placements out more widely as of May. While we've been tracking how ads work, tests and usability through our own pilot and other studies, we have not had a view into how often they show up, what types of prompts they’re shown for, and how well they align with those responses that they are served under.

That changed this week when ads started appearing in our AI tracker.

Study: Content Recency's Impact on AI Visibility in 2026

Admin — Fri, 24 Jul 2026 14:27:35 GMT

Does recency still matter in AI search? How much?

The New Analytics Blind Spot: AI Traffic & AI Agents That Shop Without Visiting Your Site

Admin — Wed, 22 Jul 2026 04:00:00 GMT

This blog was originally published on March 3, 2025 and updated on July 22, 2026.

Is That "Web Visitor" an Agent or a Human?

Back in January 2025, OpenAI’s Operator became the first true "AI agent" to hit the market, available to ChatGPT Pro users. We ran an early experiment to see how it would show up in GA4, and the results were strange enough that we wrote about them in the original version of this blog.

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

Admin — Fri, 17 Jul 2026 22:43:50 GMT

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 to a peak of 1,195 in a matter of weeks. The page was also cited in a Google AI Overview.

Why The Best Strategy for Creator Ads Focuses on Testing and Speed

Admin — Thu, 16 Jul 2026 16:14:10 GMT

TL; DR

Brands investing in creator ads may not be taking the right actions to see their desired results
The industry is focused on helping brands find the right creator, when the real problem is about inconsistent performance
One Seer client saw lower-funnel ROAS vary from 0.69x to 20x and customer acquisition costs spanning $23 to $538
A cost gap will emerge between brands optimizing for creator selection and brands who build a testing framework

Yes, Everyone Is Doing Creator Ads

Almost every paid social team we talk to is putting money into creator ads right now. Most of them will be let down by the results, but not for the reason they might think.

Culture: What things we got rid of, changed, kept and why!

Admin — Wed, 15 Jul 2026 22:21:02 GMT

The Case of the Missing Click: What Zero-Click Search Actually Means for Paid Strategy

Admin — Tue, 14 Jul 2026 16:24:27 GMT

Who doesn’t love a good mystery? I've been thinking about the conundrum that is zero-click search a lot lately, partly because of Rand Fishkin's recent study showing that less than one-third of Google searches still send a click. But it’s also on my mind because of what I keep hearing from marketing teams across the industry.