The 3-Month AEO Plan for a New SaaS Company

A founder emailed us last month. Their product is a Series A workflow tool with $4M ARR, decent SEO, and zero AI presence. They ran a scan: 0 mentions in ChatGPT across 40 relevant prompts. Their two biggest competitors averaged 12 and 17. That gap is what the next 90 days is about closing.

AEO is not SEO with a new coat of paint. The mechanics rhyme, but the surface area is smaller, the signals are weirder, and the feedback loops are slower. Google will rank a page in 3 days. ChatGPT might take 6 weeks to reliably surface the same brand, and it won't tell you why. So the plan below is built around what actually moves the needle for a small SaaS: high-density content assets, structured data, distributed citations, and measurement you can trust.

This is a 12-week plan. It assumes one dedicated marketer, one part-time engineer for schema work, and a founder willing to be quoted. If you have less, cut scope. If you have more, don't accelerate — the crawl and index cycles on LLMs won't reward you for it.

Weeks 1-2: Baseline, then map the prompts that matter

You cannot improve what you haven't measured. Before you write a single blog post, you need to know two things: what prompts your buyers are actually asking AI, and where you currently show up.

Start with the prompts. Not "best CRM software" — that's an SEO keyword. Think in conversational form. "What's a good Notion alternative for engineering teams?" "Which HR tools integrate with Rippling?" "Best invoicing tool for a solo consultant in Europe?" You want 60-100 of these. Get them from three sources: your sales team's discovery calls, your support tickets, and G2/Reddit threads where your product category comes up.

Now baseline. Run each prompt through ChatGPT, Claude, Gemini, and Perplexity. Record:

Whether you're mentioned
Position in the list (first, second, buried)
Which competitors are mentioned
Which sources the AI cites (this is gold)

That last bullet is the most important. If Perplexity keeps citing a specific G2 category page, a Reddit thread, or a specific comparison article, you know where the model is pulling its opinions from. Those are your target citation surfaces. Being cited by the sources that AI trusts is worth more than any single piece of content you'll publish.

By end of week 2, you should have a spreadsheet with 60-100 prompts, a baseline mention rate across four AI platforms, and a list of 15-25 "trusted sources" that keep appearing in citations.

Weeks 3-4: Fix the foundation on your own site

Before publishing new content, make sure existing content is legible to LLMs. Most SaaS sites aren't. Their homepages are three-word marketing headers over videos. Their product pages describe features in first-person plural. Their FAQs live in an accordion that renders after a JS click.

Here's what to fix in weeks 3-4:

Schema markup. Add Organization, Product, and SoftwareApplication schema across the site. Add FAQPage schema to any FAQ content. Add Article schema with author and datePublished to every blog post. This is a 3-day engineering task, not a 3-week one. Don't overthink it. A real "About" page. Not a founder story. A structured page that clearly states: what the product does, who it's for, what problems it solves, what it costs, what it integrates with, and who founded the company. LLMs love this page because it disambiguates. If your brand name is "Pulse" or "Sync" or any other overloaded word, this page is how you tell the model which Pulse you are. Rewrite product pages for extraction. Every product page should answer, in plain prose, in the first 200 words: what the product is, who it's for, and the top 3 use cases. Then a comparison paragraph naming 2-3 alternatives and where you differ. LLMs pull quotes, not paragraphs, so write in extractable chunks. Kill the low-effort AI content. If you have 40 blog posts written by an "AI SEO agency" in 2023, most of them are hurting you. Google's helpful content system already downranks them. ChatGPT's crawler weights domain quality. Delete or heavily rewrite anything that reads like a listicle summarizing other people's listicles. Fewer, better pages beat volume every time.

By end of week 4, your site is technically clean, schema-marked, and free of the content debt that's suppressing your domain authority in LLM training data.

Weeks 5-6: Publish the two content types that actually work

Two formats punch above their weight in AEO: comparison pages and integration pages. Publish four of each in these two weeks.

Comparison pages are the highest-leverage asset in AEO. When someone asks Claude "Should I use Linear or Jira?", the model is drawing on comparison content that lives on both vendor sites and on third-party review sites. If you don't have a "You vs. Competitor" page, you're invisible for that entire class of prompt.

Here's the format that works:

Section	Purpose	Length
Direct answer paragraph	Answer "which one should I use?" in 3 sentences	60-80 words
Feature comparison table	Side-by-side, honest, includes where competitor wins	15-25 rows
Pricing comparison	Real numbers, not "contact sales"	100 words
Best-for section	"Choose X if..." and "Choose Y if..."	200 words
Migration notes	How to switch, if applicable	150 words
FAQ with schema	6-10 questions	400 words

Be honest about where the competitor wins. LLMs weight balanced content higher than promotional content. A comparison page that says "Ahrefs has a deeper backlink index than us, but we're faster for content ideation" reads as trustworthy. A page that claims total superiority reads as marketing and gets discounted.

Integration pages are the second workhorse. For every meaningful integration you support, publish a page: "Foxish + Slack," "Foxish + HubSpot," etc. Each page needs an explanation of the integration, a setup walkthrough, and a use case. This does two things: it ranks for long-tail queries ("does X integrate with Y?"), and it tells LLMs the ecosystem you belong to. When a user asks "recommend a tool that works with Salesforce," models pattern-match on these pages.

By end of week 6: 4 comparison pages + 4 integration pages live, schema-marked, linked from the main nav.

Weeks 7-8: Case studies with real numbers and Reddit presence

A case study without numbers is a testimonial with extra words. In these two weeks, publish three case studies and start seeding Reddit and Hacker News authentically.

The case studies need three things LLMs love: specific customer names, specific metrics, and specific timeframes. "Acme Corp cut their onboarding time from 14 days to 3" beats "customers see dramatic improvements." Add Review and Organization schema to make the quotes machine-readable.

Format each case study with a bolded lead answer (what changed and by how much), then a short "before" section, then "what they did," then "the result," then a founder or ops-lead quote. Cap at 900 words. Nobody's reading a 2,500-word case study, including the AI.

Now the harder part: Reddit and community presence. The single most-cited third-party source across AI platforms is Reddit. Perplexity cites it constantly. ChatGPT's search mode pulls from it. Claude references it. If your category has an active subreddit and you're not participating, you're missing the largest citation surface that exists.

Rules for Reddit that keep you from getting banned and actually work:

Comment 20 times before you post
Never link to your product in a top-level post
When someone asks for recommendations and you fit, respond as a founder ("Founder of X here, we do this...") and include a competitor honest-comparison
Answer questions about your category even when your product doesn't fit
Post one real, high-effort thread per week — a teardown, a data share, a lessons-learned

Founders always want to skip this. They say "Reddit is a distraction, we should be publishing content." Then their competitor with 400 Reddit karma gets recommended by Perplexity twice as often. The math is not subtle.

Weeks 9-10: Third-party surfaces and podcast citations

By now your own site is doing its job. The remaining gap is off-site signal. Weeks 9-10 are about showing up on the sources that LLMs already trust.

Priority list, in order:

G2, Capterra, TrustRadius. Claim your profile if you haven't. Get 15 real reviews in 60 days by emailing every happy customer with a direct link. Category pages on these sites are aggressively cited by AI.

Product Hunt. If you haven't launched, launch. If you have, get on relevant collections. The permanence of a PH page in AI training data is underrated.

Founder podcasts. Get on three. Not the biggest ones — the tightly-scoped category ones. A 45-minute conversation on a mid-tier podcast becomes a transcript, which becomes indexed content, which becomes training material. One founder we work with went from 3 to 22 monthly ChatGPT mentions after a run of 4 podcast appearances in a niche category.

Guest posts on category authorities. Not link-farm SEO guest posts. Actual expertise pieces on sites that show up in your citation research from week 1.

Wikipedia. If your company is notable enough (Series A with press coverage often qualifies), a well-sourced Wikipedia entry is the single most-weighted source in most LLM outputs. Don't create it yourself — get a third party to draft it and submit through proper channels. A rejected page is worse than no page.

The best AEO strategy is being the answer everyone else quotes.

Weeks 11-12: Measure, iterate, double down

You're now 10 weeks in. Time to look at what worked and pour gas on it. Re-run the baseline scan from week 1. You should see movement on at least 30-40% of prompts. If you don't, one of three things is happening:

Your prompts were too broad. "Best CRM" is impossibly competitive. "Best CRM for a 5-person B2B agency" is where you can actually win.
Your content is technically fine but topically thin. LLMs need repetition across sources to build confidence. One comparison page isn't enough — you need the page, plus a Reddit thread, plus a G2 review, plus a podcast mention.
You're in a category dominated by legacy players. In some categories, breaking in takes 6-9 months, not 90 days. Don't panic; keep publishing.

In weeks 11-12, identify your five most-improved prompts and your five most-stubborn ones. For the improved ones: publish supporting content that reinforces the pattern. For the stubborn ones: figure out what source the AI keeps citing and get yourself into that source.

Weekly metrics to track from here on:

Mention rate across your 60-100 prompts (target: 25%+ by end of month 3)
Position when mentioned (first three, or buried?)
Citation share — of prompts where AI links to a source, how often is it yours?
Referral traffic from AI platforms — small but growing, and directionally telling

What not to do

A short list of shortcuts that will actively hurt you:

Do not publish AI-generated content at volume. Google's March 2024 core update crushed sites that did this, and there's evidence LLM crawlers deprioritize domains with detectable AI-generated slop. If you're going to use AI in your writing process, use it as a first draft that a human heavily edits. Not as a publish button.

Do not keyword-stuff for AEO. LLMs don't parse keyword density; they parse meaning. Ten mentions of "best invoicing software" on a page won't help you. One clear, honest paragraph explaining who your invoicing software is for will.

Do not fake reviews. G2 and Capterra flag them. Reddit users notice them. And when an LLM cites a fake review, it's citing something that will eventually get taken down, taking your mention with it.

Do not obsess over one platform. Founders check ChatGPT daily and ignore Perplexity because Perplexity has less traffic. But Perplexity users convert at 3-5x the rate of Google Search users — they're pre-qualified and further down the funnel. Rank there too.

Do not expect linear progress. AEO gains are lumpy. You'll go three weeks with nothing, then a training data refresh will move 20 prompts at once. Trust the process, keep the weekly rituals.

The reality after 90 days

If you execute this plan, at day 90 you'll have: 8 comparison and integration pages, 3 case studies with schema, 15+ G2 reviews, a Reddit presence with real karma, 3-4 podcast appearances, and clean technical foundations. You'll be mentioned in 25-40% of your target prompts across major AI platforms. You won't be beating category leaders, but you'll be in the conversation.

That's the goal at 90 days. Being in the conversation. From there, month 4-6 is about compounding the wins and pushing mention position from "listed third" to "listed first."

Measuring all of this is where most teams fall over. You need to be running the same 60-100 prompts every week across four platforms, tracking mention rate, position, and citations, and correlating that with your referral traffic. Doing this manually takes about a day per week. Foxish does it automatically, tracks the deltas, and shows you which competitors are gaining or losing ground on which prompts — so you know where to spend the next week's effort. Start a free scan →. First scan is free, no card required.