226: MOST BRANDS ARE CHASING AI VISIBILITY BACKWARDS WITH ALISA SCHARF, CHIEF AI OFFICER AT SEER INTERACTIVE
ALISA SCHARF
Seer Interactive
Chief AI Officer at Seer Interactive, where she runs the agency's AI practice and researches how language models perceive and recommend brands.
A client walks into Seer Interactive asking how to win at AI visibility. Alisa Scharf's first question back is what does that even mean to you, and who's asking? Alisa is Chief AI Officer at Seer, running the AI practice across every client account, and she has the data to explain why the question matters. Her team ran more than 15,000 prompts across over half a million pages. The websites pulling the most ChatGPT citations had domain authority in the 20-to-40 range, not the 80-to-100 every tool has told brands to chase for a decade.
Citations, Alisa argues, are a leading indicator, not a result. Closer to ranking on page two of Google than to a sale. A study her colleague Sani Vasquez led found that large language models recommend a specific brand only 2.3% of the time. That number sounds broken until you see the competitive set averages about the same, which is the moment it turns from a failure into a benchmark.
Alisa's prescription starts with defense, not offense. Before chasing category terms, run a brand accuracy audit: pick factual prompts about your company, its founding date, location, what it sells, who it competes with, and find where ChatGPT, Claude, and Gemini get you wrong. Fix the record first. The harder problem comes after visibility. There's a real gap between getting cited in an answer and getting an agent to buy, book, or provision on a customer's behalf, the Supabase pattern where Claude Code picks the database and sets up the account before the human knows what Postgres is. Most tools sell the first thing and call it the second. The tooling for the second barely exists.
KEY TAKEAWAYS
- Run a brand accuracy audit before chasing category terms. Pick five factual prompts about your company (founding date, location, what you sell, who you compete with) and run them across ChatGPT, Claude, and Gemini, then fix what the models get wrong before spending on visibility.
- Treat citations as a leading indicator, not a business outcome. Citations swing month to month and model to model, so real impact shows up in direct traffic, branded search, and brand recognition, none of which sit cleanly on a dashboard.
- Separate AI search visibility from agent readiness. Getting cited in an answer and getting an agent to actually buy, book, or provision are two different problems, and most tools that claim to measure the second only measure the first.
- Check whether AI traffic self-identifies in your log files. Many agents do not, so diagnostic work to identify which bots reach your website, and what your rules let them do, is the first concrete step toward agent readiness.
- Write a canonical brand identity and complete your Organization schema. Decide what is objectively true about your company, its founding, location, and services, then audit where models hallucinate and correct the source they are reading.
SHOW NOTES
Defense Wins Championships
Offense wins games. Defense wins championships. Alisa Scharf applies that to AI visibility, where the unglamorous work of correcting what models already believe about your brand beats swinging for category terms you don't yet deserve to rank for. That work rarely produces an immediate scoreboard bump, which is exactly why most teams skip it.
The Brand Accuracy Audit
Seer Interactive builds what Alisa calls a brand accuracy audit, and the method is deliberately boring. You write a list of objective, factual prompts, when the company was founded, where it's based, what it sells, who it competes with, then run them across the models your audience actually uses. SparkToro helps confirm whether that audience lives in ChatGPT, Claude, or Gemini.
For each model, you track what it consistently gets right and what it consistently gets wrong. Seer ran the audit on itself and found the models still tie the agency to a San Diego office it no longer has. The models also failed to credit Seer for its proprietary data infrastructure, millions of dollars of investment that never made it into the story on Seer's own website or on third-party websites.
That last category is where the real gains hide. When the wrong information sits on a page you control, you can fix it today. The hardest brand problems are the ones where the model is right and your website is silent.
Why Low-Authority Websites Out-Cite the Giants
One Seer study ran more than 15,000 prompts across over half a million pages. The pages earning the most citations skewed toward low domain authority, the 20-to-40 band, not the 80-to-100 names every legacy tool rewards. Alisa compares it to the early internet: niche pages nobody would visit as a buyer somehow signal value inside the model's training data.
Her bet is that this won't last. Once enough people notice that answers lean on an obscure "agencyspotter" page instead of Search Engine Land, the credibility pressure forces a cleanup. So Seer treats strange citations as weather, not climate. Check next month and the top source is back to Reddit or YouTube.
Citations Are Page Two of Google
Alisa is blunt about the metric everyone wants on a dashboard. A citation means your page was referenced. It does not mean your brand was mentioned, and it definitely doesn't mean you were recommended. She maps a maturity model that runs citation, then mention, then recommendation, and a study led by her colleague Sani Vasquez found models recommend a specific brand only 2.3% of the time.
So what should sit on the front page of the report instead? Direct traffic. Branded search. The customer who solved a problem with your content and remembered where the answer came from. None of those connect neatly to AI search in an attribution model, which is the uncomfortable part. Alisa's view is that AI search behaves more like a branding channel than a performance one, and the teams still grading it on referral traffic are measuring 10 to 20% of the real story.
Visibility Is Not Readiness
Alisa and host Slobodan Manic kept circling one gap. Getting found is GEO. Getting an agent to complete a task on your website is something else, and the tooling for it is, in her word, catastrophic. Submit your website to most "agent readiness" tools and you get an AI search visibility report wearing a different label.
The pattern already runs in developer tools. A vibe coder in Claude Code gets told to use Supabase, says go ahead, and the agent provisions the account. The human never learns what Postgres is because the human was never the customer. The agent was.
That pattern is leaving tech faster than the tooling is catching up. If an agent might one day buy, book, or provision on your customer's behalf, the bots need your inventory, shipping, and pricing in a form they can act on, and many agents don't even self-identify in your log files yet. Diagnose that first.
The Foundation Most Brands Skipped
Alisa's catch-up advice for a 15-year-old brand is older than AI: do the positioning work. Plenty of enterprises, asked for their unique value proposition, return corporate jargon and no real answer. Models behave like matchmakers, and a smaller, sharper competitor that says exactly who it serves will beat the conglomerate that claims to serve everyone.
WATCH ON YOUTUBE
QUESTIONS ANSWERED
What is a brand accuracy audit for AI search?
A brand accuracy audit, as defined by Seer Interactive's Alisa Scharf, is a list of objective factual prompts about a company (when it was founded, where it's based, what it sells, who it competes with) run across ChatGPT, Claude, and Gemini to find what each model consistently gets right and wrong. The brand then corrects the inaccurate information, prioritizing pages it controls. Alisa recommends doing this defensive work before spending on ranking for competitive 'best X for Y' category terms.
Do low domain authority websites get cited more by ChatGPT?
In one Seer Interactive study of more than 15,000 prompts across over half a million pages, websites with domain authority in the 20-to-40 range earned more ChatGPT citations than high-authority sites in the 80-to-100 range. Alisa Scharf attributes this to how volatile the citation space is and expects models to clean up their citation sources over time, so she advises against over-optimizing for it.
How often do LLMs recommend a specific brand?
A Seer Interactive study led by Sani Vasquez found that large language models recommend a specific brand only 2.3% of the time. Alisa Scharf notes this sounds alarmingly low until you see that competitive sets average roughly the same, which reframes the number as a benchmark rather than a failure. A model is far more likely to cite a page or mention a brand than to issue a direct recommendation.
What is the difference between AI search visibility and agent readiness?
AI search visibility means getting cited or mentioned in an AI answer, which falls under GEO and AEO. Agent readiness means an AI agent can actually complete a task on your website, such as buying, booking, or provisioning on a customer's behalf. Alisa Scharf argues these are two different problems and that most tools marketed for 'agent readiness' only return AI search visibility reports.
Are AI citations a good metric to measure?
Alisa Scharf treats AI citations as a leading indicator rather than a business outcome, comparing them to a page-two Google ranking. Citations show your page was referenced but do not guarantee your brand was mentioned or recommended, and they swing significantly month to month and model to model. She recommends watching direct traffic, branded search, and brand recognition for the real business impact.
How should brands respond to losing organic traffic to AI search?
Seer Interactive saw clients lose 30 to 80% of organic traffic, and Seer itself lost about 80% over two years while revenue and leads held flat or grew. Alisa Scharf's response is to stop treating AI search as a performance channel and treat it more like a branding channel, accept that referral traffic is only 10 to 20% of the story, and shift attention toward direct traffic, branded search, LinkedIn, social, and live events.
RELATED ARTICLES
LIGHTHOUSE FAILS YOUR LLMS.TXT WITHOUT MARKDOWN LINKS
Chrome's new Lighthouse Agentic Browsing audit shipped in version 13.3.0. I ran it on nohacks.co. Five audits passed or returned not-applicable. One failed, the llms.txt audit, because my file is a `.txt` file that the parser treats as markdown, and my links were not encoded in markdown link syntax. The fix is five characters per link. The file is still served as plain text. Only the audit result changed.
A THIRD OF FINTECH IS INVISIBLE TO AI AGENTS
36% of the top fintech websites in the world deliver less than 80% of their homepage content to AI agents in the raw HTTP response. That number puts a measurement on the Structure pillar of Machine-First Architecture, the upstream requirement that every AI visibility strategy depends on. Here is what the data says and how to audit your own homepage in thirty seconds.
SEARCH AND AGENTS ARE ONE PRODUCT. YOU ONLY NEED ONE PLAYBOOK.
Sundar Pichai says Google Search is becoming an agent manager where users complete tasks, not browse links. Nick Fox says the optimization is the same for both. If you are treating search and agents as two disciplines, you are running two playbooks for one product.
ENJOYING THIS EPISODE?
Practical strategies for making your website work for AI agents and the humans using it. Read by SEOs, developers, and AI researchers. Exclusive tools, free for subscribers.