How a 200-person SaaS can outrank a $160B giant in AI search

This guide is for SaaS marketers and technical content teams at companies that will never out-spend their category leaders. The problem: large incumbents have domain authority, backlinks, and brand recognition that take decades to build. The result you'll get here: a structured GEO playbook that makes your content easier for AI engines to cite, even when you're David and your competitor is Adobe.

Prerequisites

Before you start, make sure you have:

• A functioning website with at least 20 published content pages
• Access to your CMS to edit page metadata and add structured data
• A list of 10-15 specific use-case queries your target audience asks AI engines
• A way to benchmark your current AI visibility (winek.ai measures this across ChatGPT, Perplexity, Gemini, Claude, and others)
• Basic familiarity with JSON-LD or a developer who has it

The Descript example: why this works at all

Descript is a video and podcast editing tool with roughly 200 employees. Adobe, its competitor, is worth approximately $160 billion and has thousands of engineers. CapCut is backed by ByteDance. Neither of those companies is hurting for content budget.

Yet Backlinko's SaaS LLM visibility case study found that Descript competes directly with these giants when AI engines answer questions about video editing software. The reason isn't magic or a lucky Reddit thread. It's that Descript's content is structured around specific, answerable questions in a way that makes it easy for language models to extract and cite.

AI engines are not running a popularity contest. They're running a retrieval contest. The most retrievable answer wins, not the biggest brand.

Step 1: Map the exact questions AI engines are answering in your category

What to do: Open ChatGPT, Perplexity, and Gemini. Type in 10 queries your customers actually use. Things like "best tool for removing filler words from podcast audio" or "how do I transcribe a video automatically." Write down which brands get cited and which don't.

Why it works: AI engines answer specific, bounded questions better than vague ones. If your content is optimized for broad terms like "video editing software," you're competing on the same terrain as Adobe's $50M content budget. If you own "automatic transcript editing for podcasters," you're competing on a much smaller field.

Real metric: BrightEdge research has found that AI-generated answers pull from long-tail, question-formatted content at a significantly higher rate than from category landing pages. Specificity is a structural advantage for smaller brands.

Pro tip: Pay attention to the phrasing AI engines use in their answers. Those phrases are the exact keywords you need to own in your content.

Step 2: Restructure your content around direct answers, not narratives

What to do: Take your five best-performing blog posts and audit them for answer density. Count how many direct, self-contained answers appear in the first 300 words. If the answer is zero or one, rewrite the opening to front-load the core claim.

Why it works: LLMs extract text in chunks. A chunk that starts with "Descript automatically transcribes audio in under two minutes" is citable. A chunk that starts with "When it comes to the world of content creation, many professionals find themselves..." is not. One of those is a quotable fact. The other is throat-clearing.

Real metric: A study from Search Engine Land found that AI Overviews in Google disproportionately cite pages that answer the query within the first 100 words. The same pattern holds in conversational AI engines.

Pro tip: Write a one-sentence summary at the top of every article. Not for humans. For machines. Humans will read past it. Models will index it.

Step 3: Add structured data that makes your content machine-readable

What to do: Implement FAQPage, HowTo, and SoftwareApplication schema markup on the pages most relevant to your core queries. Use JSON-LD, not Microdata. Google's structured data guidelines are the canonical reference here.

Why it works: Schema markup is the clearest signal you can send to a retrieval system. It doesn't just describe your content. It structures it in a format that models can parse without inference. A FAQPage schema tells a model exactly where the questions and answers are. You're not hoping the model finds the right chunk. You're handing it the chunk with a label.

Real metric: Moz's analysis of structured data consistently shows that pages with FAQ schema appear in AI-sourced answer boxes at higher rates than equivalent pages without it. The effect is especially strong for informational queries, which are exactly the queries AI engines handle most.

Pro tip: Don't fake the FAQ. If you write a fake FAQ just to get the schema, models will eventually penalize it through lower citation rates because the answers won't match real user queries. Use your actual customer support tickets as the source for FAQ content. That's where the real questions live.

Step 4: Build a cluster of supporting pages around your specific use cases

What to do: For every core query you identified in Step 1, create a dedicated content page. Not a paragraph. A page. Each page should be 600-1200 words, answer one question thoroughly, and link back to your main product or feature page.

Why it works: AI engines treat topic clusters as authority signals. If a model sees five pages from your domain that all address podcast transcription from different angles (workflow, accuracy, pricing, comparison, tutorial), it builds a confidence score that your brand is a legitimate source on the topic. Adobe has this for everything. You can have it for your specific niche.

Real metric: Gartner's 2024 marketing predictions estimated that by 2026, search volume via traditional engines will drop 25% as users shift to AI assistants. That means the pages being built now are the ones AI engines will have indexed and trusted when the volume shift fully arrives.

Pro tip: Use your HowTo schema on tutorial pages and your FAQPage schema on comparison pages. Match the schema type to the content intent, not just the format.

Step 5: Measure and iterate using AI visibility benchmarks

What to do: Run the same 10 queries from Step 1 through multiple AI engines every four weeks. Track which ones cite you, which ones cite your competitors, and how the language in the citations changes. This is exactly what winek.ai is built to measure.

Why it works: GEO without measurement is just content marketing with extra steps. The feedback loop between what you publish and what gets cited is the entire game. A 200-person company can't win a volume war. It can win an iteration war.

Pro tip: When a competitor gets cited and you don't, read the exact passage the AI pulled. That tells you what structural or content element you're missing. Copy the structure, not the content.

Quick reference: step summary

Step	Action	Effort	Impact
1	Map AI-answered queries in your category	Low	High
2	Rewrite content to front-load direct answers	Medium	High
3	Add JSON-LD schema (FAQ, HowTo, SoftwareApplication)	Medium	High
4	Build use-case content clusters	High	Very High
5	Measure AI citations and iterate monthly	Low	Very High

Scoring criteria for impact ratings

| Rating | Definition | |---|---|| | Low | Marginal lift, visible in 6+ months | | Medium | Measurable lift in 2-4 months | | High | Measurable lift in 4-8 weeks | | Very High | Compounding effect, builds citation authority over time |

Common mistakes to avoid

• Writing for readers only: Content that flows beautifully for humans but buries its core claim in paragraph four will not get cited. Models don't read for pleasure. They extract.
• Using schema as decoration: Adding FAQPage schema to a page that doesn't have real Q&A content will eventually hurt you. Models cross-reference the schema with the content. Mismatches reduce trust scores.
• Targeting category-level queries first: "Best video editing software" is Adobe's query. You don't have the domain authority to compete there yet. Own the niche queries first and work outward.
• Publishing once and moving on: AI engines update their retrieval indexes. A page that gets cited today may not get cited in three months if a competitor publishes something more structured. This is an ongoing process, not a one-time optimization.
• Ignoring the comparison pages: AI engines frequently cite comparison content when users ask "X vs Y" questions. If you don't have a page comparing yourself to every major competitor, you're leaving citations on the table.

Frequently asked questions

Q: Can a small SaaS company realistically compete with enterprise brands in AI search?

Yes, and the Descript example is the clearest current proof. AI engines don't weight citations by company size or marketing budget. They weight by content retrievability, specificity, and structured clarity. A 200-person company that answers specific questions well will consistently outrank a $160B company whose content is optimized for broad SEO terms.

Q: How long does it take to see results from GEO changes?

Structured data changes and answer-density rewrites can show measurable citation improvements in four to eight weeks, depending on how frequently AI engines re-index your domain. Content cluster builds take longer, typically two to four months before the authority signal compounds. Measurement cadence matters: run your benchmark queries monthly so you can see what's working.

Q: What schema types matter most for SaaS GEO?

For SaaS, the three most impactful schema types are FAQPage for informational and comparison content, HowTo for tutorial and workflow content, and SoftwareApplication for your core product pages. All three should be implemented as JSON-LD in the page head. These schema types directly map to the query formats AI engines handle most often in software research contexts.

Q: How do I know which AI engine to prioritize?

Start with Perplexity and ChatGPT because they show citations visibly, which makes your benchmark tracking easier. Gemini matters if your audience skews toward Google Workspace users. The structural content changes you make will improve performance across all engines simultaneously, because the underlying retrieval logic is similar. Don't optimize for one engine at the expense of others.

Q: Is GEO replacing SEO or running alongside it?

Running alongside it, for now. The content and structural improvements that drive AI citations, direct answers, schema markup, specific use-case pages, also improve traditional search performance. The practices reinforce each other. The divergence comes in measurement: Google Analytics tells you about clicks, winek.ai tells you about citations. You need both numbers to understand your full search presence.