Most SMBs we talk to know they should be doing something about AI visibility. They've heard the numbers. They've read the doom-scrolling LinkedIn posts about ChatGPT replacing search. But when they ask "where do I start?", the answer they usually get is a 47-step technical audit that takes a quarter to finish and assumes they have a developer on staff.

This is the simpler version. Four weeks. Four focused themes. Concrete actions you can do yourself or hand to whoever runs your website. At the end, you'll have a baseline score, a fix list, and the start of a content pipeline that compounds.

It's not the complete playbook. It is the playbook that gets you from "invisible to AI" to "starting to be cited" in a month, without a full agency engagement.

Week 1: Diagnose

Time budget: 2 hours, one sitting.

The biggest mistake SMBs make is fixing things in the wrong order. Before you touch your website, find out what's actually broken.

Action 1.1 — Run a baseline audit. Ask ChatGPT, Perplexity, and Claude the three or four queries your customers would actually use to find a business like yours. "Best [your category] in [your city]". "What's a good alternative to [your largest competitor]?". "Recommend a [your product] supplier for a small business". Note whether your brand appears, in what context, and which competitors get mentioned instead. This is your starting position.

Action 1.2 — Check the technical basics. Open three pages on a separate browser tab:

  • Your robots.txt at yoursite.com/robots.txt. Look for Disallow: / under any User-agent. Look for GPTBot, ClaudeBot, PerplexityBot, Google-Extended. If they aren't named with Allow: /, they may be getting caught by a wildcard block.
  • Your homepage source code (right-click, "View page source"). Search for application/ld+json. If there's nothing, you have no schema markup. This costs you.
  • yoursite.com/llms.txt. If it 404s, you're missing a file that costs nothing to create and tells AI tools what your business actually is.

Action 1.3 — Pick one platform to be your reference. ChatGPT has the largest user base. Make that your primary measurement surface for the next four weeks. You'll measure across all five eventually, but trying to optimize for everything at once at week one is a recipe for paralysis.

End of week 1 deliverable: a one-page snapshot. Where you currently appear (or don't), what's technically broken, and which three queries you're going to track weekly.

Week 2: Fix the foundations

Time budget: 3-4 hours, ideally one focused afternoon.

The technical work that actually moves the needle is short. Most of the SEO industry will try to sell you the long version. Resist.

Action 2.1 — Unblock AI crawlers in robots.txt. Add explicit User-agent: GPTBot (and ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot) blocks with Allow: /. If you have something to genuinely hide (admin pages, checkout flows), block those specifically. Don't use blanket Disallow: / rules. We covered the specific common mistakes in 5 Things to Fix Today to Get Cited by ChatGPT.

Action 2.2 — Add a llms.txt file. This is a plain text file at the root of your domain that tells AI tools what your business is, what you sell, your pricing, your key pages, and your contact info. It takes 20 minutes to write. There's a free generator at VisibAI's llms.txt tool if you want a starting structure.

Action 2.3 — Add the three pieces of schema that matter most. Organization schema on every page (so AI tools know what your company is). LocalBusiness schema if you serve customers locally. FAQPage schema on your most-trafficked content page. These three cover 80% of the rich-result eligibility you need. Anything more advanced is week 3 or beyond.

Action 2.4 — Audit your most important page. Pick the one page on your site that, if a buyer landed on it, you'd want them to convert. For an SMB this is usually your homepage or your main service page. Make sure it has: clear H1, a one-paragraph description of what you do at the top, your service area named explicitly, your differentiators stated plainly, and a clear CTA. AI tools synthesize answers from text that is structured and direct. Walls of marketing copy don't summarize well.

End of week 2 deliverable: robots.txt fixed, llms.txt published, three schemas added, one page tightened up.

Week 3: Create one piece of cite-worthy content

Time budget: 4-6 hours including drafting and publishing.

This is the week most SMBs skip. It's also where the long-term compounding starts.

AI tools don't just read your homepage. They synthesize answers from content across the web. The content that gets cited has three characteristics: it directly answers a question your buyer would ask, it includes specific data or examples, and it's structured so an LLM can extract a single paragraph as a citation.

Action 3.1 — Pick one query. From your week 1 list, pick the single highest-value question your buyer asks. Not the broadest. The one most likely to convert if you got the answer.

Action 3.2 — Write a 1,000-1,500 word answer. Format it as: clear question-style title, one-paragraph direct answer at the top, then supporting detail in numbered or H2 sections. Include at least one number or specific example. Include at least one comparison ("X vs Y") if relevant. Avoid sales language at the top of the page. AI tools penalize obvious selling more than they penalize plain language.

Action 3.3 — Publish, link to it from your homepage, submit to Google Search Console. Make sure the new page has its own canonical URL and is in your sitemap.

This one piece of content won't move the needle in a week. It moves the needle over two or three months as it gets crawled, indexed, and (if it's good) cited by the AI tools you care about.

Week 4: Measure and decide

Time budget: 1 hour for measurement, then a decision.

Action 4.1 — Re-run your three reference queries. Same three queries from week 1, same three platforms. Did anything change? Note: probably not yet. AI tools don't update overnight. The week 4 baseline is for tracking three months from now, not for celebrating in week 4.

Action 4.2 — Decide what's next. You now have a real choice. Either:

  • Continue the sprint cadence on your own. Repeat weeks 3 and 4 monthly. Build out one piece of cite-worthy content per month. Track quarterly. This works for SMBs with internal content capacity.
  • Bring in tooling or expertise. If you've done weeks 1 and 2 and the technical bar is now beyond your in-house skill, AI visibility tools (like ours, but the principle holds across the category) can measure cross-platform visibility, surface specific fixes, and shorten your iteration loop. This is where most SMBs land after a sprint.
  • Do nothing for now and revisit in six months. A legitimate choice if AI search isn't yet a meaningful share of your buyer's journey. But the cost of starting late grows over time, because content compounds and the brands writing answers in 2026 will dominate the citations in 2028.

The point of the sprint is that none of these decisions are speculative anymore. You've done the actual work, you've measured the actual gaps, and you can make the choice with real information.

What this playbook deliberately doesn't include

It doesn't include detailed schema implementation for niche industries. It doesn't include a competitor analysis framework. It doesn't include the per-platform optimization differences between ChatGPT and Perplexity and Claude. Those are real and worth doing, but they're week 5 onwards, not the foundation.

The mistake most SMBs make is starting with advanced tactics before fixing the foundations. Four weeks of unglamorous foundation work beats six months of clever tactics on a broken base.

Start with the audit this week. Get the baseline. The rest follows.

If you want a faster version of week 1 with a real score and a specific fix list, you can run a free AI visibility audit here in about 8 minutes. Either way, the work is the same. The sooner you know your starting position, the sooner the four weeks start counting.