LaunchRanked

Playbook · AI products

SEO & AI search playbook for AI startups

AI product categories fill up fast. A new task (“turn a podcast into clips”, “write product descriptions”) can have dozens of tools within months, and most of their sites say the same things: fast, easy, powered by the latest model.

What separates pages in this category is evidence. Real inputs and outputs, the models you actually use, honest limits and clear costs. It's also the category where founders are most tempted to publish thousands of AI-written pages at once, which is how we learned about indexing throttles on our own product.

Updated By the LaunchRanked team7 sources checked

Searches to plan around

Query patterns and the intent behind them. We don’t quote search volumes: check your own Search Console for the ones you already appear for.

  • “AI [task] tool”, “AI tool to [verb] [object]”

    Find a tool for a job

    The main discovery pattern. One page per task you really do well, showing a real before and after.

  • “free AI [task]”

    Try before paying

    Say plainly what the free tier includes. Vague “free” claims lose the click to a clearer competitor.

  • “[model] vs [model] for [task]”

    Technical comparison

    Searched by people choosing between tools built on different models. Only write these for models you run.

  • “[tool] alternatives”, “[tool] pricing”

    Switching and cost

    Credit-based pricing confuses people. Pages that work out the real cost per output get read.

  • “how much does [output] cost with AI”

    Budgeting

    Cost pages were one of the families we built for Versely. Use your own pricing and state your assumptions.

Page families that work

  • Task pages with real outputs

    /tools/podcast-to-clips

    Show a real input, the output your product made, and what it can't do. The examples make each page different from every competitor's.

  • Model pages

    /models/[model-name]

    Which model powers which feature, what it's good at in your product, and when you switched. Update them when models change.

  • Audience pages

    /for/real-estate-agents

    Only where the audience uses a different workflow or output format. Show their example, not a generic one.

  • Alternatives, X vs Y and cost pages

    /alternatives/[competitor]

    Fact tables with prices, output limits and model choices, dated so readers know when they were true.

Risky page types

  • “AI for [every profession]” pages

    Hundreds of pages with the same body and a new noun are what Google's scaled content abuse policy describes, and its generative AI guidance says the same about generating many pages without adding value.

  • Galleries of user outputs

    Public share pages for every generation create huge numbers of thin URLs. Curate a small gallery with context and noindex the rest.

Technical pitfalls

  1. Publishing too much, too fast

    When we published about 2,500 posts in one month on Versely, indexing got throttled. A steady pace, around 20 a day at most for that site, worked better. Start slower on a newer domain.

  2. The product is the page

    Many AI tools are a single-page app where the tool, the pricing and the explanation all render client-side. Put the explanation, examples and FAQ in server-rendered HTML around the app.

  3. Stale model names

    Pages that name a model you stopped using mislead readers and assistants. Keep model names in one data file so every page updates together.

How AI assistants answer questions in this category

When someone asks an assistant for “the best AI tool for [task]”, the answer is usually a shortlist built from AI tool directories, listicles and comparison pages that the assistant retrieves. Google describes a similar process for AI Overviews and AI Mode as query fan-out, where one question becomes several related searches. Your own site gets cited more when it answers the sub-questions directly: price, free tier, output format, supported languages, which model.

What tends to earn a citation

  • A one-sentence description of the task you do, used the same way on your site and in every directory listing.
  • Pages with a real example output and the exact settings that produced it.
  • Plain pricing: cost per output or per credit, and what the free tier gives you.
  • Allowing the AI search crawlers you want to appear in, such as OAI-SearchBot and PerplexityBot, in robots.txt.

Check whether the AI crawlers you want can reach your site with our AI crawler checker.

  • AI tool directories

    There are many, and quality varies a lot. Start with maintained, reviewed ones from our AI tool directories list and check each before paying for a featured slot.

  • Launch platforms

    A launch gives you a spike and a permanent listing. Launch free on LaunchRanked for a followed link after review, with no badge required.

  • Newsletters that cover AI tools

    Many AI newsletters pick tools every issue. A concrete example output is a better pitch than a feature list.

A 90-day plan

  1. Days 1–30

    Foundation

    • Server-render the explanation, pricing and FAQ around your app.
    • Allow the AI search crawlers you want in robots.txt and publish an llms.txt if you like.
    • Set up Search Console and write down your top tasks by customer usage.
  2. Days 31–60

    Task and model pages

    • Publish one page per core task with a real example output.
    • Publish model pages for the models you actually use.
    • Submit to a shortlist of AI tool directories and launch once.
  3. Days 61–90

    Comparisons and the loop

    • Add alternatives and cost pages with dated fact tables.
    • Rewrite titles on pages at positions 5–20.
    • Noindex near-duplicate pages Search Console shows as competing.

What to watch in Search Console

The Performance report gives clicks, impressions, CTR and average position by query, page, country, device and date. Traffic from Google's AI features is counted there too, under the Web search type (Google).

Indexed vs submitted pages
If the Page indexing report shows many “Discovered – currently not indexed” URLs, you're publishing faster than Google wants to crawl. Slow down.
Impressions on task pages
Task queries are your discovery channel. Rising impressions with low clicks usually means the title needs work.
Queries containing model names
Shows which model comparisons people expect you to answer.

Frequently asked questions

Can we use AI to write our SEO pages?

Yes, if each page adds something real and is checked for accuracy. Google's guidance is about value and accuracy, not the tool used. Mass-generating near-identical pages is the problem.

How fast should a new AI startup publish?

Slower than you think. On our own product, publishing about 2,500 posts in one month got indexing throttled; around 20 a day was a safe pace for that site. Newer domains should start lower.

Do we need llms.txt to be cited by AI?

Google says you don't need any special AI file for AI Overviews or AI Mode. llms.txt is a proposal some tools read; it's cheap to add but not a requirement.

Which directories are worth it for AI tools?

Maintained ones with a review step and real traffic. Check each site before paying; many AI directories launched recently and vary a lot in quality.

Sources

Checked on September 23, 2026. Search patterns and page advice are our own judgement from running this playbook; we don’t quote search volumes or third-party statistics.

  1. Google Search Central: Spam policies for Google web search
  2. Google Search Central: Guidance on using generative AI content
  3. Google Search Central: AI features and your website
  4. OpenAI: Overview of OpenAI crawlers
  5. Perplexity: Perplexity crawlers
  6. llmstxt.org: The /llms.txt proposal
  7. Search Console Help: Performance report
  • AI agents

    Show real runs, explain permissions plainly, and make your own site readable by the agents that browse it.

  • Creator tools

    Platform guides and spec pages kept current, and audience pages built around real creator workflows.

  • Indie hackers

    A few hours a week, aimed at niche searches the big players ignore, with launches that leave something behind.

All SEO playbooks by startup type