Skip to content

GEO

AI Search Engine Optimization:
A Do-It-Yourself Guide

A practical AI search engine optimization plan for SaaS founders doing it themselves: baseline, crawler access, content, outside mentions, and tracking.

Published · 8 min read

GEO ai-search-optimizationgeodiy

You have checked ChatGPT a few times, your competitor shows up and you do not, and the agency quotes you got cost more than your monthly marketing budget. AI search engine optimization is something a solo founder can do in-house: most of the work is writing clear pages, making sure crawlers can read them, and getting other sites to describe you accurately. This guide is the step-by-step version, in the order I would do it on a small SaaS with no marketing hire.

If you have already decided to pay someone, the guide to AI search optimization services covers costs and how to vet a provider. This article is for the case where you are doing the work yourself.

For a side-by-side look at what the agency lists agree on, what retainers cost, and what to ask before signing, see how to judge the best answer engine optimization services.

What AI search engine optimization actually means

AI search optimization (you will also see it called AEO or GEO) is the practice of getting your brand mentioned, cited and described accurately in answers from ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews. HubSpot defines it in almost exactly those terms, and the tool vendors use the same definition. If you came here from the marketing side, our guide to what AEO means in marketing turns the same idea into a four-step weekly process.

Google’s position is worth reading before you start. Its Search Central guide on AI optimization says that, from Google’s side, optimizing for generative AI search is still SEO, and that AEO and GEO are just names people use for work aimed at AI search experiences. Its AI features run on the core ranking and quality systems, with retrieval-augmented generation and query fan-out layered on top: the engine splits one user question into several related searches, pulls pages for each, and writes an answer from what it retrieved.

That has two practical consequences for a DIY plan:

  • Your existing SEO work is not wasted. Strong SEO foundations support AI visibility, and HubSpot, MonsterInsights and Google all describe the two as complementary.
  • Ranking alone is not enough. According to an Ahrefs study, 38% of AI Overview citations came from pages within Google’s top 10 results. MonsterInsights reports a different cut of the data: about 46% of documents linked in AI Overviews came from top organic results for exact-match queries. Either way, a large share of cited pages sit outside the top results, which means the engines pull from a wider pool than your rank tracker shows.

Step 1: Write down your baseline before changing anything

Skip this and you will never know whether anything you did worked. Spend an hour on it.

  1. List 10 to 20 questions a buyer would type before they find you. Mix category questions (“best invoicing tool for freelancers”), comparison questions (“X vs Y”), and problem questions (“how do I stop chasing late payments”).
  2. Run each one in ChatGPT, Perplexity, Gemini and Claude. Use a logged-out or fresh session where you can, so personal history does not skew the answer.
  3. For each answer, record three things in a spreadsheet: whether you are mentioned, which competitors are mentioned, and which URLs are cited.
  4. Note anything wrong: outdated pricing, a feature you killed two years ago, a confused description of what you do.

The cited URLs column is the most useful part. It tells you which pages the engines trust for your category, and those pages are your outreach list for step 4.

Step 2: Make sure AI crawlers can read your site

This is the cheapest fix on the list and the one founders most often get wrong without knowing it. Some hosting setups, security plugins and CDN bot settings block AI crawlers by default.

Check these in order:

  • Open your robots.txt and look for rules that disallow GPTBot, ClaudeBot, PerplexityBot or Google-Extended. The reference guide on how to allow AI crawlers in robots.txt lists the user agents and the exact lines to add.
  • Check your CDN or firewall bot rules. A “block AI scrapers” toggle can override a permissive robots.txt.
  • Make sure your key pages render their main text in HTML. If your pricing page or feature list only appears after JavaScript runs, some crawlers will see an empty page.
  • Keep your sitemap current so new pages are discovered quickly.

None of this will get you mentioned on its own. It removes the reason you could not be mentioned even with perfect content.

Step 3: Rewrite the pages engines quote

Google’s guidance here is short and specific: create valuable, non-commodity content with a unique point of view, avoid recycling common knowledge, and organize it with clear headings and structure. For a SaaS, the pages that matter most are fewer than you think.

PageWhat engines look forDIY fix
HomepageA one-sentence answer to “what is this and who is it for”Put the category and the audience in the first paragraph, in plain words
PricingCurrent, readable numbersText, not an image; same numbers everywhere on the site
Comparison pagesHonest side-by-side factsOne page per real competitor, with a table and where each tool fits better
Use-case pagesA specific job done for a specific userName the user, the problem and the result
FAQ or docsDirect answers to direct questionsQuestion as heading, answer in the first two sentences

A few rules I follow when rewriting:

  • Lead with the answer. If the heading is a question, the first sentence answers it.
  • Use specific facts that are easy to quote: limits, integrations, prices, setup time. “Supports Stripe and Paddle, set up in about ten minutes” gets quoted. “Powerful integrations” does not.
  • Write the opinion only you can write. If you built the product because existing tools failed at something, say what and why. This is the non-commodity part Google is asking for.
  • Do not spin up dozens of near-duplicate pages for every phrasing of a query. Google explicitly warns against creating separate content for every query variation to manipulate rankings or AI responses, and query fan-out already covers the variations for you.

Add structured data (Organization, Product, FAQPage) once the copy is right. Markup helps engines parse a page, but it cannot fix a page that does not say anything specific.

Step 4: Get mentioned on pages you do not own

When you ask an assistant “what is the best tool for X”, it rarely relies on your homepage alone. It leans on lists, reviews, comparison articles and forum threads. Your baseline spreadsheet from step 1 already tells you which of those pages the engines cite for your category.

Work through that list:

  1. For each cited “best tools” article where you are missing, contact the author with a short note: what you do, who it is for, one concrete differentiator, and a link to a page with your current pricing. Most will ignore you. Some will update the list.
  2. Claim and complete your profiles on review and directory sites your category uses. Outdated profiles are a common source of wrong answers about pricing and features.
  3. Answer real questions in communities where your buyers ask for recommendations, under your own name, disclosing that you built the product.
  4. Publish something other people want to cite: a small dataset from your own product, a benchmark, a clear teardown of a common problem.

This step is slow, and it is the one most people stop doing after two weeks. One outreach attempt a week, every week, does more than ten in one afternoon.

Step 5: Track results and decide what to do next

After four to six weeks, rerun your baseline prompts. You are looking for movement on specific prompts, not a general feeling that things improved. Did a comparison page you published start getting cited? Did an engine stop quoting your old price? Did a competitor overtake you on a category prompt?

Manual reruns work for a while. They get painful once you have 20 or more prompts across four engines, and the variation between runs makes week-to-week comparisons noisy. The walkthrough on how to track your brand in ChatGPT covers the manual method in detail, including how to log results so they stay comparable.

When you want this automated, the tools range widely in price. According to Contentpen’s roundup, OtterlyAI starts at $29/month and Profound at $99/month, and HubSpot lists its AEO tool at $50/month ($45/month paid annually). AskAiRank has a free tier with 10 prompts on 2 engines polled once a day, and the Starter plan at $19/month covers 3 brands, 25 prompts and 3 engines at a time. Pick whatever fits; the important part is that the same prompts run on a schedule so the numbers mean something over time.

A realistic first month

If you want a concrete schedule, this is what I would do with about five hours a week:

  • Week 1: baseline spreadsheet, robots.txt and CDN check, fix anything blocking crawlers.
  • Week 2: rewrite the homepage opening and pricing page; make sure prices match everywhere.
  • Week 3: publish one comparison page against the competitor that shows up most in your baseline.
  • Week 4: send outreach to the three most-cited “best tools” pages that leave you out, and update your directory profiles.

Then rerun the baseline at the end of week 6 and compare. Start today with the spreadsheet: ten buyer questions, four engines, three columns. Everything else in this guide depends on having that list.

Frequently asked questions

How many hours a week does DIY AI search engine optimization take?

The first pass (baseline, robots.txt check, fixing your homepage and pricing copy) fits into one or two focused days. After that, plan on a few hours a week: one new or rewritten page, one outreach attempt for an outside mention, and a short review of your prompt results. Consistency over a few months matters more than a big push in week one.

Do I need separate pages for every question people ask ChatGPT?

No. Google's own guidance warns against creating separate content for every query variation just to manipulate rankings or AI responses. One strong page per real buyer question, with clear headings and specific facts, covers the variations that engines generate when they expand a query.

Can I just ask ChatGPT about my product myself instead of using a tracker?

For the first baseline, yes, and it is a good exercise. The problem is that answers vary between runs, accounts and engines, so a handful of manual checks can mislead you. Once you have more than ten prompts or want to see trends week to week, a scheduled tracker saves time and gives you numbers you can compare.

If my Google rankings are good, am I already covered in AI search?

Partly. Ahrefs found that 38% of AI Overview citations came from pages in Google's top 10, so ranking helps but does not decide everything. Chat assistants also lean on third-party pages like review sites, comparisons and forum threads, so you need outside mentions on top of your own rankings.

When does it make sense to stop doing this myself and hire someone?

When you know exactly what needs to happen but cannot find the hours, for example you need ten comparison pages written and nobody on the team can write them. If you are still unsure what is wrong, hiring early tends to buy activity rather than results. Get your own baseline first, then decide.

Founder working late at three monitors

Track your AI visibility

See how your SaaS appears in ChatGPT, Claude, and Perplexity.

3-day trial, no card: 10 prompts on ChatGPT and DeepSeek, checked daily. Then from $19 a month.