Get Cited in ChatGPT and Perplexity Without an Agency

Key Takeaways

• Agencies selling "GEO" or "AI SEO" services are mostly repackaging two things you can do yourself: getting mentioned on high-citation third-party platforms, and writing direct-answer content instead of marketing pages.
• In this category specifically, Reddit accounts for roughly half of all AI citations, more than every other source combined. That's not an agency service, it's genuine participation.
• The technical basics (robots.txt allowing AI crawlers, an llms.txt file, server-rendered pages) take under an hour to set up correctly and most agencies charge for exactly this.
• The actual bottleneck for most small SaaS teams isn't expertise, it's consistency: showing up in threads and publishing direct-answer content on a schedule, not a one-time push.
• Budget for one thing an agency is genuinely faster at if you're going to spend anywhere: earning a real editorial mention (one or two pieces, not many) on a site with existing AI-citation weight in your category.

---

What agencies are actually selling you

Most "AI visibility" or "GEO" agency retainers break down into three deliverables: technical setup (llms.txt, schema, crawler access), content written to be more directly citable, and outreach to get you mentioned on third-party sites. None of these require specialized tooling you can't access yourself, they require time and consistency, which is genuinely what you're paying for when you hire someone.

If you're a solo founder or small team with more time than budget, here's the same playbook done yourself. If you haven't already, it's worth understanding why AI engines skip brands in the first place, since that's what every step below is actually working against.

---

Step 1: Fix the technical basics (one afternoon)

Confirm your robots.txt explicitly allows the AI crawlers: GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot. Many hosting setups block unknown user-agents by default via a WAF, which silently kills your AI visibility while your Google SEO looks fine, because Googlebot is usually allowed and these aren't.

Add an llms.txt file at your root domain, a plain-text summary of what your product does and links to your key pages. It's a lightweight, emerging standard, not a guaranteed citation lever, but it costs almost nothing to add and there's no reason to skip it.

Confirm your key pages render real content without JavaScript. Check by viewing your page source (not the rendered DOM) for a product page, if the actual text isn't there, a crawler that doesn't execute JS won't see it either.

---

Step 2: Write direct answers, not product pages

This is the part most self-serve attempts get wrong. Writing "10 Reasons to Choose [Your Product]" doesn't get cited, because it isn't the answer to anything a buyer actually asked. Writing "How to tell if a competitor is winning your category in ChatGPT" is a direct answer to a real question, and that's the shape of content that gets pulled into AI synthesis.

Build a list of the actual buyer questions in your category (see how to find which competitors ChatGPT recommends over you for exactly how to build that list), and write one focused, specific, opinionated post per question. Mention your product where it's a genuine part of the answer, not as a forced pitch.

---

Step 3: Show up where citations actually concentrate

This is the step that's easy to skip because it doesn't feel like "content work," and it's the highest-leverage one. Find the real, active threads in your category on Reddit (search your category name plus "reddit," look for recent, active subreddits, not dead ones) and answer questions genuinely, the way you'd help a friend, with your product mentioned only when it's actually the honest answer.

Do the same on relevant G2/Capterra category pages (claim your listing, get real customer reviews, don't skip this because it feels unglamorous, review sites are cited constantly), and on LinkedIn, where posts with concrete numbers and specific claims get cited more than generic thought-leadership.

None of this scales through a single push. It compounds through weekly consistency, which is the actual thing an agency retainer buys you: someone doing this every week whether or not you feel like it that week.

---

The one thing worth paying for, if you pay for anything

If budget allows for exactly one paid lever, it's earning a mention in an existing, high-authority editorial piece in your category (a "best tools for X" roundup on a site that's already cited a lot, not a random guest post). One or two of these carry outsized weight, we've seen single articles account for double-digit percentages of a category's total AI citations. That's worth a genuine PR push even on a tight budget. Everything else in this playbook is free, just not fast.

---

Track whether any of it is working

The hardest part of doing this yourself is knowing if it's landing, since AI answers change and you're not going to manually re-run 20 prompts across 5 engines every week. GetIntel automates exactly that check: it runs your real buyer prompts on a schedule and shows you which fixes actually correlated with a new citation, so you're not guessing whether the Reddit comment from three weeks ago did anything.

Start tracking your AI citations free and see where you stand before you spend a dollar on outside help.