Sources: Ahrefs study of 75,000 brands and Ahrefs llms.txt study of 137,000 domains, plus Originality.ai adoption tracking. All general findings. Note the second number carefully. We implement llms.txt and we are telling you upfront that almost nobody reads it, because an agency that sells you that file as a strategy is selling you a placebo.
The Right Fit
- The brand is newer, less recognized, or has inconsistent facts about it floating around the web. Different descriptions, different categories, name collisions.
- The goal is being the name a model volunteers unprompted, not just the source it cites for one specific question.
- There is a real commitment to building third party authority over months, the same way every durable competitive advantage gets built.
- The technical and content foundation is already in motion, since GEO’s entity and authority signals compound fastest on top of it.
What’s Included
Entity Clarity and Organization Schema
Structured data and consistent brand facts across your site so models can retrieve unambiguous information about who you are. Across 75,000 brands, branded mention frequency correlates with AI citation at 0.664, against 0.218 for backlinks. It is one of the strongest levers a brand actually controls.
llms.txt Implementation
A plain language summary file at your site root that gives AI crawlers a direct map of your most important pages. We will set it up, and we will also tell you the truth about it: adoption has passed 36,000 sites, but 97% of these files receive no requests at all. It is a cheap baseline signal, not a strategy, and anyone pitching it as the centerpiece of a GEO engagement is selling you something that does not work.
Authority and Citation Building
Getting your brand mentioned on the third party sources models already trust. This is the slow part and it is the part that actually moves the needle, which is why it starts on day one rather than after the quick technical work is done.
Entity Consistency Audit
Checking that your name, description, category and core facts match everywhere they appear. The failure modes here are unglamorous and expensive. On one client site we found two versions of the homepage indexed separately, splitting the authority on the single most valuable URL the brand had. On another we found an old renamed URL still indexed and quietly resolving to the homepage, bleeding 169 impressions at zero clicks. Neither shows up in a content review.
Topical Depth Around Your Brand
A single page rarely earns a model’s trust on its own. Building real depth around your core offerings gives generative engines many consistent, corroborating signals to draw from instead of one isolated claim.
Cross Platform Visibility Tracking
Checking whether ChatGPT, Claude, Gemini, Grok, and Perplexity actually surface and recommend your brand for the queries that matter to you. We run the prompts ourselves and log what comes back, because no platform reports this.
How This Gets Built
- Clarify. Lock down one consistent name, description, and category everywhere the brand appears.
- Structure. Add Organization schema and an llms.txt file as the baseline technical signal.
- Earn. Build mentions on the third party sources models already weight heavily.
- Deepen. Surround the core offering with real topical content, since a single page rarely earns a model’s trust alone.
- Check. Test real prompts across every major platform on a recurring basis, since citation share moves.
Why Buyers Now Ask AI Who to Trust
Your next client may never scroll a list of links. They ask a model who the best option is and act on the answer. Each number below is labeled by the kind of evidence behind it.
Sources: 6sense Buyer Experience Report (roughly 4,000 respondents), Adobe Digital Insights Q2 2026 AI traffic report, NP Digital analysis of 60 client campaigns. Two caveats we are not going to bury: Adobe’s figure comes from US retail sites, so read it as directional if you are a service business, and Adobe published it alongside its own AI optimization product. In the September 2026 review we also removed a Gartner forecast that search volume would fall 25% by 2026. It was made in 2024, we are now in 2026, and Google still holds about 91% of search. A prediction that has come due does not get to keep being quoted as a prediction.
What This Looked Like on a Real Build
A destination race with no history, no backlink profile, and no brand recognition outside its own city. Exactly the profile GEO is supposed to fix. Today, ask ChatGPT, Claude, Gemini, Grok, or Perplexity which marathons to run near St. Petersburg or Tampa and the race comes back in the answer on all five. That is an internal observation from our own prompt testing, not a number any platform hands you, and we check it on a schedule rather than once.
The search side, over the same period. Client results from that client’s own Google Search Console and Race Roster accounts, June 9 to September 8, 2026.
Over the same window, registration pace ran 540% ahead of the four months before the work, with the launch week burst stripped out of the math. Correlation, stated as correlation. Pricing, email, and the race’s own community move registrations too, and we are not going to pretend we can separate them cleanly.
Disclosure: Scott Fishman is Executive Director of the St. Pete Marathon as well as Director of Growth at Winsignia, which is how we have account level access to this data. Worth knowing when you read it.
St. Pete Marathon
The full write up of the build behind those numbers, including what was broken at the start.
United Sports Association
A donor focused homepage for a Florida nonprofit, built to earn trust and drive support.
Entity Authority Compounds. So Does the Gap.
A smaller competitor building clearer entity signals and real third party mentions gets recommended by ChatGPT or Claude instead of you, no matter how much better the business actually is. A bigger, messier competitor is just as beatable. GEO rewards clarity and consistency over size.
The entity and authority work built today is what a model draws on months from now. That is why starting earlier wins the recommendation later.
Scott Fishman is Director of Growth at Winsignia, where he leads the AI search optimization work behind every build on this site. Reach him on LinkedIn.
FAQ
We are the market leader. Why would AI not already recommend us?
Market position and AI recommendation run on different signals. Being the biggest does not guarantee strong entity clarity or a clean third party footprint. A smaller, more clearly defined competitor will outperform a bigger name on exactly those signals.
What is GEO?
Generative Engine Optimization is the discipline of shaping how large language models like ChatGPT, Claude, and Gemini decide which brands to mention when generating an original answer, based on entity clarity, authority, and structured technical signals.
How is GEO different from AEO?
AEO targets citations inside a live answer engine response. GEO targets being recommended by a model that may not be actively searching the live web at all, drawing instead on what it already learned about your brand or retrieved from trusted sources.
Does adding an llms.txt file actually get us cited more?
No, not on its own, and we would rather say that plainly than sell it. Ahrefs studied 137,000 domains and found 97% of llms.txt files receive zero requests of any kind. Google has also stated the file is not required to appear in AI features. We implement it because it costs almost nothing and may matter later, not because it does anything today.
How long until GEO shows results?
Technical signals like schema and llms.txt can go live in days. Model level recommendation strength builds over months as authority and consistent mentions accumulate. It compounds rather than resets, which is the argument for starting before you feel ready.
What do the first 90 days look like?
An entity consistency audit and the fixes it surfaces, Organization schema and llms.txt live on the site, the first third party mention targets identified with outreach started, and a documented baseline of what ChatGPT, Claude, Gemini, Grok, and Perplexity currently say about you. That baseline is what makes every later gain measurable instead of anecdotal.
Can you guarantee we will get recommended?
No. Anyone who does is guessing at behavior they do not control, on systems that get retrained without notice. What we can commit to is a documented baseline, the specific entity and authority work that moves those signals, and honest reporting of what changed, including when something did not work.
Does GEO work for a small or newer brand?
Yes, though it takes longer to build the branded mention volume that correlates most strongly with citation. Focusing on a narrow, well defined niche with real depth behind it is usually the fastest path for smaller brands. The client build described above started with no brand recognition outside its own city.
Want to Be the Brand AI Recommends?
Book a call and we’ll map what it takes to get there for your business.