Are You Invisible to AI Search? A Playbook for B2B Small Businesses
Your buyers ask ChatGPT, Google AI Mode, Perplexity, and Claude for vendor shortlists, often before they ever open your website. Here is how to find out whether you show up, and exactly how to fix it if you do not.
A CMO I spoke with had invested heavily in SEO and ranked number one for her primary keywords. Traffic looked healthy. Then she ran the same buyer questions through ChatGPT and Perplexity, and her brand never appeared. Not once. Competitors with weaker SEO were cited constantly. "We thought we had search figured out," she said. "We optimized for the wrong engine."
That scene is repeating across thousands of B2B companies. The way buyers find vendors has changed, and most small and mid-sized firms have not measured the gap yet.
This is your practical playbook to close it. No agency required. No large budget. A clear process, a spreadsheet, and about five minutes to start.
What actually changed, and why it matters to you
For two decades, "being findable online" meant ranking on Google. You earned a spot on page one, a buyer clicked your link. That model still exists, and it just stopped being the whole story.
Today a growing share of B2B buyers open ChatGPT, Perplexity, Claude, or Google AI Mode and ask: "Who are the best [your service] providers for a company like mine?" The engine returns a short list. They contact those companies. If you are not on the list, you were never in the conversation.
This practice has a name, GEO (generative engine optimization), sometimes called AEO (answer engine optimization). The distinction is simple: SEO earns a link in a list, GEO earns a mention inside the answer the AI writes.
The shift is measurable. Gartner projects traditional search volume will fall about 25% by 2026 as buyers move to AI assistants. A randomized, pre-registered field experiment by researchers at the Indian School of Business and Carnegie Mellon University, run in January and February 2026, found Google AI Overviews cut organic clicks by 38% on the queries where they appear, with zero-click searches rising from 54% to 72%. The Pew Research Center, analyzing 68,879 searches, found people click a result about 8% of the time when an AI summary is shown, versus 15% without it. And where Google once returned ten blue links, AI engines cite only a handful of sources per answer. The room is smaller, and the seats are assigned by the model.
Human trust became a retrieval signal in 2026
One 2026 change is worth understanding on its own, because it shows where this is heading. Google built a feature called Preferred Sources in stages. It launched in the United States and India in August 2025, expanded to English worldwide in December 2025, and opened to all languages on April 30, 2026. On May 27, 2026, Google extended it into AI Overviews and AI Mode, with rollout through June 2026, and the same announcement introduced a Highly Cited badge for original, widely referenced reporting.
The feature lets a person open their Search settings and hand-pick the sites they want to see more of. Once they do, those sites carry a visible Preferred badge inside AI Overviews and AI Mode, not only in Top Stories. Today the badge appears when one of your chosen sites already shows up in a response. Google has said it is working on using Preferred Sources as a ranking signal across its AI features, so chosen sites surface more often. Google reports that people click through to a preferred source about twice as often, though that figure is Google's own and has not been independently verified.
The mechanic is the part worth sitting with. Every other lever in search is something you optimize toward until an algorithm rewards it. This one has no setting to tune and no score to chase. The only input is a human deciding you are worth choosing. That is the principle our Agent Found service is built on: a business earns machine visibility by being clear, useful, and trusted enough that both people and the systems reading on their behalf select it.
For an owner-led B2B company, this is not a campaign to run. Your buyers are not setting source preferences for a machine shop or a sign fabricator. The signal to read from it is strategic. The platforms are now placing human choice directly into what their AI surfaces, which is the direction this playbook has been preparing you for.
Phase 1: run your 5-minute baseline audit
Before you fix anything, find out where you stand. This is a free, manual test you can run right now. The one rule: always test with non-branded questions. Every engine already knows your company name. What matters is whether it recommends you to a buyer who has never heard of you.
Build your 10-question prompt list
Think like a buyer who does not know you exist. Write 10 prompts mixing these types:
- Category and use case: "Best [your service] for [your customer type] in [your region]"
- Problem-first: "How do companies solve [the problem you fix]?"
- Buyer guidance: "What should I look for when choosing a [your category] vendor?"
- Shortlist: "Who are the top [your category] providers for small businesses?"
- Voice or conversational: "Who would you recommend for [your service] near [your metro]?"
Write all 10 in a spreadsheet first. You need the exact same prompts across every platform to compare results week over week.
Test across all four platforms
Free accounts are fine for the baseline. Spend 10 to 15 minutes running your 10 prompts on each:
ChatGPT
About 900 million weekly users (OpenAI, February 2026). With search on, it reads the live web, so your indexable, crawlable pages and your presence on third-party sites both feed it.
Google AI Mode and AI Overviews
Now the default-leading experience after Google I/O 2026, powered by Gemini 3.5 Flash. Watch the AI answer at the top and note which sources it names. AI Overviews appear in roughly a quarter of US searches.
Perplexity
Shows numbered citations with every answer. The most useful platform for seeing exactly which sources the model trusts about your category.
Claude
Enable web search. Claude leans on consensus across sources, so third-party mentions, directories, and consistent facts about your business carry extra weight here.
Score your results
For each prompt on each platform, record: does your business appear, are you named, is your site linked, and what is the tone around the mention. Then score yourself.
| Your visibility | What it means | Priority |
|---|---|---|
| 0 to 2 platforms, fewer than 3 prompts trigger a mention | Effectively invisible. Buyers researching your category will not find you in AI search. | Fix immediately |
| 2 to 3 platforms, 3 to 6 prompts trigger a mention | Partial visibility. You appear for some queries and miss most buying-intent searches. | Improve now |
| 3 to 4 platforms, 7 to 10 prompts trigger a mention | Strong foundation. Shift focus to citation quality and sentiment, not just raw presence. | Optimize |
Track your score weekly in the same spreadsheet. Models and indexes update constantly, so a brand can gain or lose AI visibility within a few weeks.
Phase 2: diagnose why you are not showing up
If you scored low, one or more of these root causes is almost certainly responsible. The fix varies by platform, because each draws from a different source.
- AI bots are blocked at the door. Many sites block AI crawlers in
robots.txtwithout realizing it. Fresh June 2026 network data reveals that OpenAI’sGPTBothas officially overtaken all other models, driving 11.48% of all global AI crawler traffic as ChatGPT climbs toward 1 billion weekly users. If your firewall or hosting provider automatically blocks aggressive crawlers, you are actively turning away the single most powerful indexing engine currently building B2B vendor shortlists. - No machine-readable signals. Without Organization, LocalBusiness, or Service schema, the data layer that engines parse cannot confirm who you are, what you do, and where you operate.
- Fluffy copy acts as a data filter. AI engines do not understand corporate vanity. Adjectives like "industry-leading," "synergistic," or "world-class" are completely ignored by scrapers because they lack structural data. If your services page doesn't state your exact parameters, capacities, or specs in plain, technical prose, the engine simply skips you to avoid hallucinating.
- No third-party mentions. If your own website is the only source about you, the model has nothing to cross-check. A Yext study of 6.8 million citations found 86% of AI citations come from brand-managed sources, split roughly 44% first-party site and 42% listings, reviews, and directories.
- Stale or thin content. A 400-word About page from 2019 will not earn a citation. FAQ and how-to content with clear, dated answers outperform static brochure pages.
Phase 3: the fix-it roadmap, ordered by impact
These fixes are ordered by speed to impact. Complete Week 1 before anything else. It is free, it takes hours, and it directly feeds the engines that read the live web.
Claim your digital real estate, free and high-impact
- Audit your firewalls and JS payloads. Beyond checking your
robots.txt, look at your security layers (like Cloudflare or Sucuri). Ensure security rules aren't flagging legitimate AI crawlers likeGPTBotorPerplexityBotas malicious scraping attempts. Additionally, ensure your core product or service specifications are rendered directly in the HTML source, not lazy-loaded via client-side JavaScript that bots ignore. - Google Business Profile. Claim, verify, and complete every field. Google AI answers pull from this directly.
- Bing Places. Claim your listing and submit your sitemap to Bing Webmaster Tools. Microsoft Copilot and parts of the AI ecosystem still read Bing.
- Industry directories. Get listed on 3 to 5 reputable directories for your sector: G2, Capterra, or Clutch for software, and the equivalent for your trade. These rank among the most-cited sources across engines.
- Consistent NAP data. Your name, address, and phone must match exactly across every listing, your site, and your schema. Inconsistency lowers an engine's confidence and it may drop you to avoid stating something wrong.
Rewrite your core pages for AI extraction
- Answer-first structure. Open every section with a direct statement of fact. Instead of "Our team has decades of combined experience," write "We provide commercial HVAC installation for buildings up to 100,000 sq ft across the Chicago metro, with 4-hour emergency response."
- Lead with evidence. The peer-reviewed GEO study found that adding statistics, source citations, and quotations produced the largest gains in AI visibility. Put real numbers and named sources into your best pages.
- Build a real FAQ page. Write 15 to 20 questions your buyers actually ask. Keep answers in the 40 to 60 word range, which sits well for extraction, and add FAQPage schema.
- Publish one category-definition guide. A thorough "What is [your service] and how do you choose a provider" piece earns citations every time someone asks an engine about your space. This is your most durable AI asset.
- Show credentialed authorship and a visible last-updated date. A named author with verifiable experience, and fresh dates on time-sensitive pages, both raise trust signals that engines weight.
Add schema markup, the machine-readable layer
Schema tells engines what your content means, not just what it says. Use JSON-LD. If you run WordPress, plugins like Yoast or Rank Math handle most of it without code.
- Organization schema for your name, URL, logo, phone, address, and social profiles. This is the entity signal engines check first.
- LocalBusiness schema for any physical location or service area, with serviceArea, hours, and a plain-English description.
- FAQPage schema on your FAQ page, which directly enables question-and-answer extraction.
- Service or Product schema on each offering page.
- Isolate and link your corporate entity graph. Use the
sameAsattribute in your JSON-LD schema to explicitly map your website to your precise corporate nodes: your exact LinkedIn Company page, your official state business registry entry, and your active executive profiles. This creates an unshakeable data triangle that tells LLMs your small-to-midsize business is a verified, real-world operation.
Validate every change at the Google Rich Results Test. One caution worth stating plainly: Google's own guidance says you do not need special AI files or markup tricks to appear in its generative results. Schema that mirrors your visible content helps. Schema that contradicts the page can hurt. Accurate, accessible content does the heavy lifting.
Build third-party authority, the hardest to fake
- Request reviews. Google, G2, Clutch. A short post-project email asking a happy client to review you on one platform takes 30 seconds and feeds AI credibility. Aim for 10 or more per platform.
- Earn local or trade press. A mention in your business journal, a trade publication, or an industry blog is worth more to an engine than ten pages of your own copy.
- Write contributed articles. A 600-word expert piece on an industry site, with your company named and linked, builds citation authority directly.
- Post consistently on LinkedIn. It is heavily read by the major engines. Two to three posts a month from the company page and from leadership builds a visible body of expertise.
- Answer real questions on Reddit and Quora. Substantive expert answers in your niche can earn citations, Perplexity especially.
Phase 4: build a tracking system that lasts
A one-time audit is a starting point, not a strategy. Models update constantly, and the sources cited in AI answers shift month to month. You need a repeatable system, and you need to measure the right thing.
Free DIY tracking, 30 minutes a week
Add these columns to your spreadsheet and run your 10-prompt test across all four platforms once a week:
Date | Platform | Prompt | Mentioned (Y/N) | Cited or linked (Y/N) | Sentiment | Competitor who appeared insteadCalculate a weekly Brand Mention Rate: the share of prompts where you appear across platforms. Watch that number, plus your branded search volume, more than raw website clicks. Clicks are falling structurally as answers replace links, so citations and mentions are the truer signal now. Note which competitors keep appearing where you do not, then study what they do differently.
One encouraging point for the effort: being cited in an AI answer does not cannibalize your traditional traffic. Industry analysis from BrightEdge found that brands cited in AI Overviews saw higher click-through on their adjacent organic and paid listings. AI visibility and SEO compound rather than compete.
Paid tools when you are ready to scale
Google Search Console
Watch for queries with high impressions and low clicks. That pattern often means your content is feeding an AI answer without earning the click. As of June 3, 2026, Search Console also has a dedicated Search generative AI report that isolates how often your pages appear inside AI Overviews and AI Mode. It shows impressions but not clicks yet, and it is rolling out to a subset of sites first, so check whether yours has it.
Citation trackers
Tools such as Profound, Semrush, and Otterly monitor where your brand appears across ChatGPT, Perplexity, and Google AI answers, with share-of-voice reporting.
Prompt and rank trackers
Platforms that map your target buyer questions to AI visibility, similar to a traditional rank tracker, suit teams already comfortable with SEO tooling.
Be your own customer
Before buying anything, query the engines yourself and log which sources get cited. That manual baseline is what every paid tool is automating.
For most small and mid-sized firms, the free spreadsheet method covers the first 90 days. Add automation once you have a baseline and are publishing fresh content each month.
Sample prompts: copy and test right now
Replace the bracketed placeholders with your details, then paste each one into ChatGPT, Google AI Mode, Perplexity, and Claude. These mirror how real buyers search.
Who are the top [your service, e.g. IT managed service providers] for small businesses in [your city or region]?What is the best way to find a reliable [your service, e.g. commercial cleaning company] for a 50-person office?What should I look for when choosing a [your category, e.g. B2B logistics partner], and who are some reputable options?Can you recommend a [your service] that specializes in [your niche] near [your city]?The 90-day action plan at a glance
| Timeline | Action | Impact |
|---|---|---|
| Day 1 | Run the 5-minute baseline audit across all four platforms | Baseline |
| Week 1 | Confirm AI crawlers can reach you, then claim Google Business Profile and Bing Places | High |
| Week 1 to 2 | Get listed on 3 to 5 relevant industry directories | High |
| Month 1 | Rewrite service pages answer-first, add an FAQ page, lead with evidence | Medium to high |
| Month 1 | Add Organization, LocalBusiness, and FAQPage schema | Medium to high |
| Month 2 | Publish a comprehensive category-definition guide | Medium |
| Month 2 to 3 | Earn 10 or more reviews per platform and secure 1 to 2 press mentions | Medium |
| Ongoing | Run the weekly 30-minute prompt audit and track your Brand Mention Rate | Essential |
Frequently asked questions
What is GEO, and how is it different from SEO?
SEO focuses on winning a ranked hyperlink in a traditional search engine results page. GEO and AEO are about securing direct citations inside the dynamic, synthesized responses generated by engines like Google AI Mode, ChatGPT, and Perplexity. The objective pivots from driving raw web clicks to earning authoritative brand recommendations within the AI context layer.
How do I check whether AI search recommends my business?
Write 10 non-branded buyer questions about your category, then ask each one in ChatGPT, Google AI Mode, Perplexity, and Claude. Record whether you are named, cited, or linked. Repeat weekly with the same prompts to track your mention rate over time.
Does schema markup actually help AI visibility?
It helps when it mirrors visible page content and reinforces your entity. The peer-reviewed GEO study found statistics, citations, and quotations lift visibility most. Google states you do not need special AI files for its generative results, so prioritize accurate, accessible content over markup tricks.
Do I need an llms.txt file to be found by AI?
Not for Google. Google Search Central states that llms.txt and similar files are not required to appear in its generative search. The file is cheap to add and some operators keep one, and it is no substitute for clear content, valid schema, and pages that are not blocked to AI bots.
How long until AI search visibility improves?
On-page work like claiming listings, fixing crawlability, and rewriting pages can show movement within 30 to 60 days. Off-page authority from reviews, press, and directories compounds over two to three months. Treat it as a measured system with a baseline and weekly retests.
The bottom line
"Being findable" no longer means only ranking on Google. It means showing up in the AI answer your buyer receives before they ever open a browser tab. Google made that the default at I/O 2026, so the question is no longer whether this affects you.
The encouraging part for small and mid-sized B2B firms: the window is still open, and early movers have a real edge. Most enterprise teams have launched their AI visibility work. Most SMBs have not started. The businesses winning right now are simply the ones who began measuring six months ago. They built a baseline, found their gaps, and improved on a schedule. That work starts with a spreadsheet and 10 honest prompts. Everything else follows from knowing where you stand.
Next step
If you want help running your baseline audit or building an AI visibility strategy for your specific business, our Agent Found service structures your B2B digital presence so AI engines can find, understand, and recommend you. It includes a full diagnostic, schema implementation, and ongoing monitoring. Book a discovery call and we will assess where you stand and show you exactly what to change.
Related resources
Agent Found, AI Search Optimization
How AI Agents Find Businesses on the Web
AI Strategy Workshop
Revenue Leak Audit
Follow Get 'er Done on LinkedIn