How to Get Cited by AI Search (and Why Google Rankings Aren't Enough Anymore)

Last week a friend who runs a 40-person SaaS told me their organic traffic dropped 28% year over year. Their rankings did not move. Their content output did not slow down. What changed was the result page.

When he searched his own brand on Google, the AI Overview answered the question in three sentences and cited two other companies. The traditional blue links were still there. Almost nobody scrolled past the summary.

That story is now everywhere. ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews are reading your content, summarizing it, and answering the user's question without sending them to your site. If you are still measuring SEO purely by Google rank position, you are measuring the wrong thing.

This is the playbook I wish someone had handed me a year ago. It covers what Answer Engine Optimization actually is, the seven tactics that meaningfully change whether you get cited, and how to track it without burning a quarter on guesswork.

The Zero-Click Decade Just Got Worse

For ten years SEOs have grumbled about featured snippets eating clicks. Snippets were the warm-up act. AI search is the headliner.

Three things have changed at the same time:

1. Google AI Overviews are now the default for most informational queries. The classic ten blue links still load, but the answer sits above them, ends with a citation list, and visually wins the attention contest before the user even scrolls.
2. People are leaving Google entirely for some questions. ChatGPT and Perplexity get used for product research, comparison shopping, and how-to questions that used to start at google.com. If your brand is not in those answers, the user does not get a second chance to find you.
3. The model picks who gets credit. Traditional SEO was a ranking game with a public scoreboard. AI citation is a synthesis game where the model decides whose information becomes the answer and whose name shows up in the source link.

Classic SEO optimizes pages to rank for queries. AI search optimizes brands to be quoted in answers. Those are related but different problems, and the second one needs its own playbook.

What "AI Search Visibility" Actually Means

You will see a few competing names: Answer Engine Optimization (AEO), Generative Engine Optimization (GEO), AI SEO, LLM SEO. They all describe the same thing: getting your brand, facts, and content surfaced inside answers generated by large language models, instead of just inside ranked search results.

Concretely, AI search visibility is measurable along three axes:

• Citation frequency. How often your domain shows up as a source link when an AI answers a relevant query.
• Share of voice. When users ask about your category, what percentage of the time does the model mention your brand by name versus a competitor's?
• Sentiment. When the model does mention you, is the framing neutral, positive, or negative? Models occasionally inherit unflattering descriptions from a single bad review site, and those descriptions stick.

None of this shows up in Google Search Console. You have to monitor it on purpose.

The Seven Tactics That Actually Move Citation Rates

I will be blunt: most "GEO tips" floating around the internet are recycled SEO advice with the word "AI" stapled on top. The tactics below are the ones that have moved citation rates on our own sites and on sites I have audited for friends. None of them require rebuilding your CMS.

1. Treat your homepage like the model's reference card

When an LLM has to describe what your company does, it grabs the first paragraph of your homepage and the meta description more often than anything else. If your hero copy says "We unlock human potential through synergistic platform thinking," the model has nothing to work with and will either skip you or generate a worse description than you wrote.

Write the homepage opening as if you were telling the model your one-line pitch: who you are, what you do, who it is for, and what makes it different. Plain English. Specific nouns. The model will quote it back almost verbatim.

2. Write a real llms.txt

An llms.txt file at the root of your domain is the AI-era equivalent of robots.txt -- but instead of telling crawlers where not to go, it tells them what is worth reading. List the URLs of your highest-signal pages (product, pricing, FAQ, documentation, your top 5 blog posts) with one-sentence descriptions of each.

Not every model fetches llms.txt yet. The ones that do are increasing fast, and the file costs you five minutes to create. Skip it and you are leaving easy ground unclaimed.

3. Use JSON-LD that names entities, not just topics

Structured data was always useful. With LLMs, it becomes a citation accelerator. The schema fields that matter most for AI search are the ones that explicitly name entities: Organization, Person, SoftwareApplication, Product, with their sameAs arrays pointing to your social profiles, GitHub, Wikipedia, Crunchbase, and so on.

Models use those links to disambiguate which "Acme" you are. The more sameAs links you provide, the more confidently the model can attach facts to your brand instead of someone with a similar name.

Look at the schema on this article. It includes a Person node for me with my portfolio, GitHub, and parent-company links in sameAs. That is not vanity. That is how the model learns that this Terrell K. Flautt is the same person as the one at github.com/terrellflautt.

4. Build FAQ schema for the exact questions people type

FAQ schema is the single highest-leverage thing most sites are not doing. When a user asks ChatGPT "is X HIPAA compliant" or "does X integrate with Salesforce," the model is hunting for a clean question/answer pair somewhere on the web. If yours is the only site that has marked up that exact question with valid FAQPage JSON-LD, you win the citation.

The discipline here is harsh: use the questions your customers actually type, not the polished ones marketing wishes they would type. Pull them from your sales call transcripts, support tickets, and the "People Also Ask" box on Google.

5. Chunk your content like a model would

LLMs do not read your blog post; they retrieve chunks of it. Long paragraphs without internal structure get split badly, which means the chunk that lands in the model's context window is often missing the topic sentence that would have made it citable.

Write with retrieval in mind. Short paragraphs. Real subheads every 150-300 words. Each section should make sense on its own if a model only pulls that one chunk. If you cannot summarize what a paragraph is about in five words, the model probably cannot either.

6. Get cited by sources LLMs already trust

This is the unglamorous tactic, and the most important. Models lean heavily on a small number of high-authority sources to anchor their answers: Wikipedia, GitHub, official documentation, Stack Overflow, established trade publications, and a handful of comparison sites per industry.

If you can be linked from those sources -- a Wikipedia entry, a "best of" roundup at a trusted outlet, a GitHub README that references your tool -- the citation transfers. Models that already trust the source will surface yours more often, and with friendlier framing.

This is also why public, signed work matters. An article like this one, published under a real person's name on a real domain, is more useful to the model than ten anonymous corporate blog posts saying the same thing.

7. Audit your share of voice quarterly

You cannot improve what you do not measure. Run the same set of category questions through ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews on a schedule, log which brands get named, and watch how that mix shifts over time.

You can do this manually with a spreadsheet. I did, for the first few months. The work is tedious. It is also where most of the actual learning happens, because you see exactly which phrasings of a question return your brand and which return a competitor's.

Where GeoGryphon Fits

The seventh tactic above -- the manual audit -- is also the one that breaks first when you have anything else to do. That is the gap we built GeoGryphon to close.

GeoGryphon is the AI search visibility tool in the SnapIT Suite. It is a sister product to Sphinx Agent, built for the same audience: founders and small teams who need this to actually work, not a $30K/year enterprise contract.

Here is how each of its core features maps back to the tactics above:

• AI Visibility Audits scan your brand across all eight engines that matter -- ChatGPT, Perplexity, Gemini, Google AI Overviews, Claude, Bing Copilot, Meta AI, and Grok -- and produce a weighted Share of Voice score with sentiment. That is tactic #7 (quarterly audit) running automatically, with the part you would have done in a spreadsheet replaced by a dashboard.
• Entity Mapping pulls the named entities the models already associate with your brand and shows you the knowledge gaps versus your competitors. It also auto-generates the JSON-LD you need to fix those gaps. That is tactic #3 (entity-rich schema) handed to you instead of hand-written at 11pm.
• GEO Content Briefs turn a target query into a citation-optimized brief in seconds: the entities to mention, the questions to answer with FAQ schema, the chunk structure the models prefer, the freshness signals to include. That is tactics #4 and #5 (FAQ schema and retrieval-friendly chunking) translated into a brief your writers can execute on without learning the theory first.
• Daily Monitoring tracks citation frequency across all eight engines so you find out about new mentions, and new omissions, the day they happen instead of the quarter they happen. That turns tactic #7 from a lagging indicator into a leading one.
• Competitor Tracking compares your citation frequency and sentiment against competitors' brands head-to-head. That is the share-of-voice axis from earlier in this article, made comparable across the players in your category.

The free tier is built to actually be useful: 3 AI visibility audits per month, 1 content brief per month, 3 engines tracked (ChatGPT, Perplexity, Gemini), no credit card. Enough to run a real audit on your own brand and one competitor before deciding whether to upgrade. Paid plans start at $29/month for Solo, $99 for Starter (adds daily monitoring), $149 for Professional (all 8 engines plus API access).

If you do nothing else after reading this article, run a free audit. You will learn more in five minutes about how AI engines describe your brand than you will from any keyword tool.

Pair It With Sphinx Agent

One more thing worth mentioning, because the two products are designed to work together. GeoGryphon helps you get found inside AI search. Sphinx Agent -- the product this blog is built around -- handles what happens next.

If GeoGryphon does its job, more of your future visitors will arrive having already read an AI-summarized version of your pitch. They will skip the homepage and ask a specific question: "do you integrate with Salesforce", "can I import from Intercom", "what is the price for the white-label tier." A static FAQ page no longer cuts it. An AI agent on your site that answers those questions in real time, captures the lead, and routes the rest to a human is now the conversion layer.

One tool to get the AI-search visitor to your door, another to make sure the conversation continues once they arrive. We use this pairing on our own sites and recommend it to friends.

Start Free

Pick one tactic from the seven above and ship it this week. The compounding starts the moment you do.

If you want a baseline before you start, run a free audit at geogryphon.com/pricing.html -- 3 AI visibility audits per month, no credit card. You will see exactly which AI engines are already mentioning your brand, which ones are not, and where the citation gaps are big enough to fix this quarter.

Then come back and read the rest of the playbook on this blog. We publish new analysis here weekly under my name, and the next few pieces dig deeper into the schema patterns, the llms.txt format, and the share-of-voice math worth tracking.