🔥 SEOs Diners Club #212: How to Measure AI Search Visibility — "Being Visible" Is No Longer Enough

Everyone's saying it: "We need to improve our visibility in AI search."

Heard it in client meetings. On LinkedIn. At conferences. Probably said it yourself.

And you're right. ChatGPT, Perplexity, and Google's AI Overviews are shaping purchasing decisions before users ever visit your website. AI search is now a critical touchpoint in the buyer journey.

But here's the question I keep asking: How many of you are actually measuring it?

This week, we dig into the KPIs that actually matter for AI search visibility — with sharp input from Aleyda Solis and Graphite CEO Ethan Smith. We also look at a finding that should shake up your entire GEO strategy: ChatGPT's default and premium models cite almost completely different sources. And there's news from Anthropic, Google Maps, and a smart move from a Turkish startup you should know about.

Grab your coffee. Let's go.

🎯 You Say You Want AI Visibility. Now Prove It.

"Being visible in AI search" is the new "we need to be on page one."

Everyone agrees. Nobody measures it.

Tomek Rudzki from Peec AI published a comprehensive guide this week on the KPIs that actually work for AI search visibility. Aleyda Solis and Graphite CEO Ethan Smith added their perspectives. When I read it, I thought: this is exactly what was missing. The "why GEO matters" conversation is largely won. The "how do we measure it" conversation is just beginning.

Here are the five metrics worth tracking — and what they actually mean for your strategy.

Visibility %: Does AI Even See You?

The most fundamental question: does your brand appear in relevant AI search responses? And if so, how often?

This is Visibility Percentage — the share of relevant AI-generated answers that include your brand. Peec AI's data shows Chime has double the AI visibility of Revolut in the US: 66% vs. 33%. One number, very different stories.

But a single visibility score doesn't tell the whole story. You need to segment your prompts by topic, by funnel stage, by customer segment. Are you appearing at awareness? Or only at decision? The difference between those two should change your entire content strategy.

Tracking individual prompts will always have noise — LLMs are non-deterministic by nature. But group prompts into categories and patterns emerge fast.

Position: Where on the List Are You?

Familiar concept for SEO professionals. The logic doesn't change in AI search. Brands mentioned earlier in an AI response attract far more attention than those buried at the end.

Think about it: if someone asks ChatGPT "best CRM software" and your brand is 10th on the list, you're getting a fraction of the attention the first or second brand receives.

LLMs don't rank brands randomly. Two factors drive position: a brand's prominence in training data, and the sources the model retrieves in real time. Both are parameters you can influence.

A Peec AI example: AI Mode placed Skoda Elroq at the top of its "best electric cars" response. For the brand, that's a visibility advantage that directly shapes purchase decisions. One position lower and that advantage disappears.

Practical note: Track position across multiple prompts and aggregate weekly. Daily results can swing significantly — weekly averages show the real trend.

Brand Sentiment: What Is the AI Actually Saying About You?

Visibility and position tell you you're in the room. Sentiment tells you what's being said about you.

This is one of the most undervalued — and most actionable — areas in AI search optimization. Training data changes slowly, but the sources shaping your brand's sentiment can often be corrected quickly. For established brands, this is often the best place to start.

When a potential customer asks "Is HubSpot easy to use?" right before making a purchase decision, they're not doing research. They're deciding. And the answer depends entirely on what AI says next.

Peec AI's finding while auditing Revolut's AI presence is striking: LLMs were consistently citing a review site (Sitejabber) filled with fake reviews. On closer inspection, 75% of the negative reviews came from single-review accounts, many promoting unrelated services. A single email from the legal team would likely fix it. Triple benefit: readers get accurate information, brand sentiment in LLMs improves, and that negative content gets less weight in future training.

This isn't a PR problem. It's a revenue problem.

Conversion & Revenue: How Do You Measure Business from AI?

It's measurable. The most practical method right now is self-reported attribution: asking customers directly how they found you.

Where you ask matters. Some businesses collect this in demo calls or during onboarding — it fits naturally in the conversation. Others ask at signup. Tally already does this: when users select "AI," Tally asks which queries they used to find the product. That tells them both their conversion rate from AI platforms and the exact prompts that drove discovery.

Once you know which customers came through LLMs, you can track the revenue they generate over time. If Google Analytics is already under-reporting organic's contribution, this method gives you a much stronger argument.

Traffic: Familiar, But Not Enough on Its Own

Traffic is the metric we reach for because it's easy to report. But AI search users rarely click. When ChatGPT recommends a CRM, there's often no link in the response. The user either searches Google next (GA4 codes it as "organic") or types the URL directly (appears as "direct"). Either way, the AI assistant gets zero credit.

Eight Oh Two's 2026 research supports this: 37% of consumers now start searches with AI rather than Google, but 85% still cross-check with traditional search before converting. One journey, two channels — and attribution models usually only catch the second one.

Aleyda Solis puts it plainly: "We need to stop using traffic as a reliable KPI. We need to shift to a mix of branding and performance KPIs: AI visibility, sentiment, purchases, and revenue."

Ethan Smith adds the statistical frame: "LLMs respond randomly, but that randomness is predictable. A sample of 10 responses is already enough for a quick estimate."

Three Steps to Start Today

1. Open up your pricing page. ChatGPT's premium model GPT-5.4 Thinking actively crawls pricing pages and has cited them 138 times. A "Contact Us" wall may be costing you visibility with premium model users. At minimum, show price ranges or a plan comparison.

2. Add self-reported attribution. Add "How did you find us?" to your signup or demo form. Make sure "AI assistant (ChatGPT, Gemini, Perplexity)" is one of the options. This will be your most reliable source of actual AI-driven conversion data.

3. Run a brand sentiment audit. Query your brand on ChatGPT, Gemini, and Perplexity. If responses are drawing from negative or inaccurate sources, correcting the source fixes both the user experience and LLM perception. Sometimes one email changes everything.

Every week you're not measuring AI search visibility is a week you can't explain why your efforts matter — to clients, to leadership, or to yourself.

⚙️ Algorithm & Search Engine Updates

Google Discover Core Update: Local Publishers Lost National Reach

According to DiscoverSnoop's analysis, the numbers are striking: Yahoo lost half its article placements, Forbes saw a 67% audience score drop. The Fox News family lost over 40% in visibility.

The most interesting finding is in local publishers. Syracuse.com dropped 80% in audience score — but when you break it down by state, the New York audience was stable. All the losses came from Florida and California. Google said it would "show more local content." But the data suggests something more specific: local publishers stayed strong in their home regions but lost national reach. YouTube placements rose 15%, continuing Google's pattern of favoring its own properties.

What it means for you: If you produce regional content, don't analyze Discover performance in aggregate. Break it down by geography. The problem may not be your content — it may be Discover's geographic filtering logic.

ChatGPT's Default and Premium Models Search the Web Completely Differently

Writesonic's analysis this week uncovered a finding that should shake up your GEO strategy: when you ask ChatGPT's default model (GPT-5.3 Instant) and its premium model (GPT-5.4 Thinking) the same question, only 7% of the cited sources overlap.

GPT-5.4 directs 56% of its citations to brand websites directly; GPT-5.3 does so only 8% of the time. More striking: 75% of the domains cited by GPT-5.4 don't appear in Google or Bing results at all. The premium model operates independently of traditional rankings.

Content preferences differ too. GPT-5.3 relies on third-party articles (Forbes, TechRadar). GPT-5.4 prefers first-party brand content: pricing pages at 19%, product pages at 10%. GPT-5.3 cited only 4 pricing pages; GPT-5.4 cited 138.

What it means for you: There is no "one size fits all" GEO strategy anymore. Third-party media coverage is critical for the free model; first-party content is decisive for the premium model. Hiding your pricing behind a "Contact Us" wall may mean you're invisible to premium users entirely.

GSC Branded/Non-Branded Filter: Now Use It

Google Search Console's branded/non-branded filter is now live for all eligible sites. Start using it immediately.

If branded traffic is rising but non-branded is falling: your brand awareness is working, but you're not reaching new audiences. Add informational content to your strategy. If non-branded is rising but branded is falling: you're being discovered, but brand loyalty is weak. Focus on conversion optimization and brand content. Weekly and monthly views are coming soon.

🤖 From the World of AI

Claude's 1 Million Token Context Window Is Now Available to Everyone

Anthropic opened Claude's 1 million token context window to general availability for Opus 4.6 and Sonnet 4.6 models. The long-context fee has been removed entirely; media limits expanded to 600 images and PDF pages. Think about what it means to analyze all your page titles and meta descriptions in a single pass during an SEO audit. We've started using this actively on client projects at Stradiji.

AI Assistants Now Account for 56% of Global Search Volume

According to Graphite CEO Ethan Smith's analysis, AI platforms are generating 45 billion sessions per month globally. ChatGPT alone represents 89% of all global AI sessions. Google's share of search-related activity fell from 89% in 2023 to 71% in Q4 2025.

SparkToro founder Rand Fishkin rightfully questioned the methodology. "Sessions" and "search queries" aren't the same thing — a single ChatGPT session might contain 20 prompts, many of them coding or writing tasks. Treat these numbers as directional. But the direction is unmistakable: AI search is growing, and the SEO vs. GEO distinction has collapsed. You need both.

Morgan Stanley: "Shocking" AI Breakthroughs Coming in H1 2026

Morgan Stanley expects major AI developments in the first half of 2026. OpenAI's GPT-5.4 Thinking model reached 83% on the GDPVal benchmark — human expert level. Small, specialized language models (SLMs) are projected to form the backbone of enterprise AI.

US Supreme Court: AI-Generated Works Cannot Hold Copyright

The Court confirmed that only human-created works are eligible for copyright protection. The "Experience" pillar of E-E-A-T just got significantly more strategic. If someone reading your content is wondering "did a human write this?" — you already have a problem.

Google Maps Gets AI Conversational Search

"Ask Maps" is rolling out with natural language queries for locations and directions. A new optimization layer is opening for local SEO. Make sure your Google Business Profile is current and complete.

🇹🇷 From Turkey: Semust Launches AI Overviews Tracking

Semust moved fast the moment Google AI Overviews arrived in Turkey. You can add your keywords directly or import them from GSC into the AI Overviews tracking tool. Built entirely in-house by Ramazan Umutlu. If you want to track your AI Overviews visibility specifically for Turkey, try it now.

💡 This Week's Action Items

1. Open your pricing page. GPT-5.4 has cited pricing pages 138 times. A "Contact Us" wall could mean zero premium model visibility. Show price ranges or a plan comparison at minimum.

2. Add self-reported attribution. Put "How did you find us?" on your signup or demo form. Include "AI assistant (ChatGPT, Gemini, Perplexity)" as an option. It's the only reliable way to measure real AI-driven conversions.

3. Run a brand sentiment audit. Query your brand on ChatGPT, Gemini, and Perplexity. If responses rely on negative or inaccurate sources, correcting those sources improves both user trust and LLM perception. Sometimes one email changes everything.

📊 Stat of the Week

7% — The source overlap between ChatGPT's default model (GPT-5.3) and premium model (GPT-5.4) when answering the same question. Same query, almost entirely different sources. One GEO strategy is no longer enough. You need to know which model prefers what — and optimize accordingly.

🔮 Next Week: Watch For

Apple's AI-powered Siri relaunch is expected with iOS 26.4 in March 2026. Gemini integration may put voice search optimization back on the agenda.

👋 Closing

This week's biggest lesson: "being visible" in AI search is no longer enough. You need to be able to measure it — and manage it.

Every week without measurement is a week you can't explain your AI search efforts to anyone. Your clients. Your leadership. Yourself.

If you want to increase your brand's visibility across both traditional search and AI platforms, Stradiji is here. Over 15 years of SEO experience, combined with GEO expertise — helping you stay found in an AI-first world.

See you next week with more developments. Take care of yourself.

Mert Erkal

Support the newsletter: if you find this content useful, buy me a coffee. ☕

About Mert Erkal Founder of Stradiji. 15+ years in SEO and GEO consulting for corporate clients globally. Author of SEOs Diners Club (English) and Dijital Pazarlama Notları (Turkish).

• Newsletter: seosdinersclub.beehiiv.com

• Twitter: twitter.com/merterkal

• LinkedIn: linkedin.com/in/merterkal

• YouTube: youtube.com/@stradiji