ChatGPT crossed 400 million weekly active users in April 2026. Here's what that means for service businesses competing for attention in conversational AI search — and what to fix this month.
Ido Cohen · Published 2026-04-12 · AI News
OpenAI confirmed in early April 2026 that ChatGPT now serves 400 million weekly active users. That is roughly the active population of the United States and the entire EU combined, and it is happening every week. For service business owners, this is no longer a "future of search" conversation. This is a now problem.
Here is what changed, why it matters for your phone ringing tomorrow, and the three things to fix before the end of the month.
400 million weekly active users puts ChatGPT at roughly half of Google's daily search audience. More importantly, ChatGPT users are not running disposable queries. They are running multi-turn research sessions: "I need a roofer in Sherman Oaks who handles tile, can do an inspection within a week, and is licensed for insurance claims — give me three options and tell me what to ask each one."
Google gives them ten blue links and an AI Overview. ChatGPT gives them a curated, opinionated answer with three names. If your business is not one of those three names, the conversation moves on without you.
Three concrete failure modes are showing up across the service-business websites we audit:
1. The site is invisible to ChatGPT's crawler. ChatGPT's web search uses OAI-SearchBot to fetch live results. If your robots.txt blocks GPTBot, OAI-SearchBot, or Bingbot (ChatGPT browses through Bing's index), you literally cannot be cited. We have seen seven-figure local service businesses with three-line robots.txt entries that quietly disqualify them from every conversational query in their market.
2. The content reads like brochure copy. ChatGPT cites pages that answer specific questions in a structured way. "We are a family-owned plumbing company with 30 years of experience serving Greater Los Angeles" gets ignored. "Sherman Oaks plumbers should expect to pay $150-$400 for a typical drain clearing call. Here are the three factors that move that price" gets cited.
3. There is no entity-level signal of authority. ChatGPT cross-references multiple sources before recommending a business. If your name shows up on Yelp and Google Business and nowhere else, you do not exist as an entity. Businesses that appear on industry directories, local press, podcast guest lists, association member pages, and review platforms outside of Google get cited far more often.
Traditional SEO is a ranking problem. You either rank #3 or #6, and the difference is incremental traffic. AI search is a binary problem. You are either in the answer or you are not in the answer. There is no #6 in a ChatGPT response. There is one, two, or three names — and if you are not there, the user closes the window.
This is also why "we rank well in Google, we don't need to worry about AI" is a dangerous read. A service business currently ranking #1 for "best HVAC company [city]" can still be entirely missing from ChatGPT's recommendation for the same query. We have measured this gap on dozens of audits. Google ranking and AI citation rate correlate, but loosely — at around 0.4 in the data we have seen.
If you do nothing else after reading this, do these:
Pull up yourdomain.com/robots.txt right now. Confirm it allows GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, and Google-Extended. If you are not sure what is in there, here is a permissive baseline that keeps you visible:
User-agent: *
Allow: /
User-agent: GPTBot
Allow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Google-Extended
Allow: /
Sitemap: https://yourdomain.com/sitemap.xml
The argument against this is "I don't want my content training their models." The reality is that for a service business, your moat is your operations, not your blog post. Citations drive booked jobs. Refusing to be cited so that a future model does not learn the phrase "we offer 24/7 emergency service" is not a strategy.
llms.txt is the emerging standard for telling AI systems what your most important pages are. Think of it as an executive summary your site hands to any LLM that fetches it. Place it at yourdomain.com/llms.txt. List the canonical pages for each service, your service-area page, your pricing or process page, and your most-cited blog content with one-sentence descriptions. Most service businesses can build this in 30 minutes. The largest and most measurable gains we have seen on AI citation rate in the last 90 days came from sites that added a clean, well-structured llms.txt.
Pick the three pages that drive the most calls today. For each one, restructure the page so the H1 is a question your customer would actually ask, the first paragraph answers it in 60-100 words in plain English, and the rest of the page expands with specifics, pricing ranges, comparisons, and FAQs. This is the single biggest lever for getting cited. ChatGPT extracts the first 60-100 words after a clear question heading more often than any other section.
A real example from a recent client: an HVAC company in Phoenix replaced "Welcome to ABC Heating and Cooling, your trusted partner for HVAC services in the Valley" with "Phoenix homeowners should expect to pay $4,800-$8,200 for a complete AC system replacement in 2026, depending on tonnage, SEER rating, and whether ductwork needs modification." Within six weeks they appeared in 14 distinct ChatGPT responses for variations of "AC replacement cost Phoenix." Their previous citation rate was zero.
ChatGPT is not the whole story. Perplexity, Claude, Gemini, and Google's AI Mode each build their answers from different blends of crawled web content, citations, and partner data. The good news is that the fundamentals overlap heavily. A site that earns ChatGPT citations almost always earns Perplexity citations. The work compounds.
The bad news is that the businesses moving on this now are pulling further ahead every week. The compounding works in two directions. Once a business is consistently cited for a category of queries in your area, it becomes harder to dislodge — the AI systems develop confidence in that source and recommend it more often, which generates more reviews and brand mentions, which reinforces the citation, which generates more reviews. Late entrants pay a price.
Two things on the horizon to keep an eye on. First, OpenAI is reportedly close to launching ad placements inside ChatGPT search results in select US categories. This will create a paid path to AI visibility for the first time and will reshape what "ranking" inside conversational AI looks like. Second, Google's AI Mode — which launched in Chrome on April 16 — is rapidly closing the gap with ChatGPT for transactional queries and will soon be the default search experience for billions of users.
Both of these mean the same thing for a service business: the surface area where customers find you is fragmenting and the rules are being rewritten in real time. The work to fix this is not exotic. It is robots.txt, llms.txt, and rewriting three pages in question-answer form. None of it requires permission. None of it requires budget. It requires that you stop treating AI search as a future problem.
400 million weekly users is a lot of conversations happening about businesses like yours. The only question is whether your name comes up in any of them.
How do I know if my service business is being cited by ChatGPT?
Open ChatGPT, switch to web search mode, and run the 10-15 queries that matter most for your business — service + city, "best [service] near me," cost queries, comparison queries. Note whether your business is named in the response and whether your website is listed in the citations. Most service businesses that have not actively worked on AI search visibility appear zero or one times in their own top 15 queries.
Is robots.txt the only thing that determines whether AI systems can cite my site?
No, but it is the gating factor. If your robots.txt blocks GPTBot or OAI-SearchBot, you are disqualified regardless of content quality. After unblocking, citation rate depends on content structure (clear question H2s with direct answers), entity signals (third-party mentions), and topical authority. Robots.txt is necessary but not sufficient.
Should I worry about ChatGPT using my content to train its models?
For a service business, the citation benefit far outweighs the training concern. Your competitive advantage is service delivery, not website copy. The customers AI sends to your phone outweigh any value of preventing model training on your service descriptions. Allow the crawlers.
How long does it take to start showing up in ChatGPT responses after fixing the basics?
Most service businesses start seeing initial citations within 4-8 weeks of fixing robots.txt, adding llms.txt, and restructuring 3-5 high-intent pages with question-answer formatting. Sustained citation across multiple queries takes 3-6 months as entity signals build.
Sources: