Why Your Business Isn’t Showing Up in ChatGPT or Google: 7 Causes, the Fix for Each, What It Costs and How Long It Takes

I asked ChatGPT one question: “Who is the best family dentist in Austin, TX?” It named four practices, then asked me a question of its own. I asked Perplexity. It named six. Then I searched Google, and its AI Overview led with a third practice. Three engines gave three different winners.

Here is the part that should worry any owner. None of ChatGPT’s four dentists had their own page on Google’s first page. That page held Yelp, Reddit, a newspaper list, two directories and the sites of three other practices. ChatGPT named its four dentists anyway.

So what decides who gets named? And if your business is missing, what should you fix first?

Most advice on this is vague. “Build authority.” “Add schema.” “Optimize for AI.” None of it tells you which problem you have. So I ran a fresh test. I asked ChatGPT and Perplexity 10 buying questions, from dentists and plumbers to bookkeepers and HR firms. I checked Google for each one. Then I opened 63 of the websites the AI named and looked for what they had in common.

The result surprised me. Some causes that guides list first explained nothing in my sample. Another cause explained most of what I saw, and it has little to do with your own website. You will find it at cause 1.

Below you get the 7 causes, ranked by how often I would expect each to explain a miss. Each one has a fix, a cost and a timeline. At the end, a 30-minute test shows which cause is yours.

Why a miss costs you customers now

Your customers already ask AI who to hire. BrightLocal’s 2026 survey of 1,002 US adults found that 45% had used AI tools for local business recommendations in the past 12 months, up from 6% a year earlier. ChatGPT led at 31%. Among people who use AI this way, 63% trust its recommendations.

Meanwhile, AI answers keep readers on the results page. Pew Research tracked 900 US adults in March 2025. They clicked a regular result link on 8% of visits that showed an AI summary, compared with 15% of visits without one. Only 1% clicked a link inside the summary. In short, buyers see names, not links. If you are not named, you are not on the list.

There is good news too. Google’s own guide says there are “no additional requirements to appear in AI Overviews or AI Mode.” A page only has to be indexed and eligible to show with a snippet. That also means no secret trick exists, so the fix is a set of ordinary jobs done well.

The test: 10 buying questions, 3 engines, 63 websites

I chose 10 questions that real buyers type. Five name a city (Austin, Texas) and five ask for a US-wide service. I ran each one in ChatGPT with web search on and in Perplexity. I also ran the same words in Google with US settings. Perplexity’s free allowance ran out after nine searches, so I could not run the HR question there.

Next, I looked up the website of every business the two engines named. After that, I read the raw page of each site, plus its robots.txt file. Here is what each engine named.

The totals: ChatGPT named 50 businesses. Perplexity named 52 across its nine answers. Across those nine questions, the two engines named 74 different businesses, and only 22 of them (30%) appeared in both lists. That number is small because the engines also disagree on who to put first.

All three engines named the same business first on only 1 of 9 questions: roofing. ChatGPT and Perplexity matched on 3. I read Google’s AI Overview for the table above and the screenshots below, but I did not count its names in the overlap totals.

One question, three winners

The dentist question shows the pattern best. Look at what each engine did with the same words.

First, ChatGPT. It said there was no single best dentist, then ranked four practices with a rating and a short reason for each.

Next, Perplexity. It led with a different practice, then put six in a table with review counts and cited two roundup sites.

Finally, Google. Its AI Overview opened with Forest Family Dentistry, and the panel beside it showed Yelp and two practice sites.

So three engines produced three winners. That is why “optimize for ChatGPT” is not one job. It is several, and the causes below map to them.

What the test showed

1. ChatGPT named businesses that Google’s first page did not show. Only 5 of the 50 businesses ChatGPT named (10%) had their own page on Google page one. For Perplexity the figure was 7 of 52 (13%). In my earlier tests the share was higher: 29% for SaaS tools and 24% for security vendors. Service businesses depend on Google page one even less.

2. The engines disagree, most of all for local businesses. For the five Austin questions, only 6 of 42 named businesses (14%) appeared in both engines’ lists. For the four US-wide questions, 16 of 32 (50%) did. In other words, a national software or service brand has a better chance of consistent mentions than a local shop.

3. Google’s first page was a directory shelf. Reddit appeared on page one for all 10 searches. Yelp appeared on all 5 Austin searches. Google also showed an AI Overview on all 10. Across the 40 organic results I read for the Austin searches, only 16 (40%) were a business’s own site. The rest were directories, forums, media lists and review sites.

4. ChatGPT questioned the buyer back. Eight of the 10 ChatGPT answers ended with a question or a form. Six of those asked about the buyer’s need. For the dentist, it asked whether I wanted care for young children or adults.

The roofing answer went further. It ended with a three-question form: type of work, property type and main priority, with a “Find the best match” button.

Two other answers, for IT and virtual assistants, asked whether I was a customer or researching competitors. That came from my account’s memory, so I do not count it as a buyer question. I mention it in the limits below.

So the engines act like a salesperson. They shortlist, then qualify. Pages that answer those follow-up questions give them something to quote.

That raises the real question. If rank on Google does not decide the shortlist, what does? To find out, I opened 63 of the named websites and checked them one by one. The answer is in the causes below.

The 7 causes, ranked by how often I would expect them

I ranked these using my test data, Google’s and OpenAI’s own documents, and my earlier studies. The ranking is my judgment, not a measured frequency. First, the map: it shows where a business can go missing.

Cause 1: Other sites do not list or review you

What I saw. The answers leaned on ratings, review counts and “best of” pages. ChatGPT’s dentist answer cited DentistsRanked. Its plumbing and HVAC answers cited ContractorsRanked. Its web design and virtual assistant answers cited Clutch, and its bookkeeping answer cited NerdWallet. Perplexity quoted review counts such as 937, 700 and 4,049 for its HVAC picks. At the same time, only 10% of ChatGPT’s picks had their own page on Google page one. So the engines were not reading the winners’ own pages to build the shortlist.

Official sources agree. Google says local prominence comes from information such as “how many websites link to your business and how many reviews you have” (Business Profile help). BrightLocal notes that AI answers may draw on reviews, directories, business websites and social media.

The fix. Open the answer, click the source chips and write down which list sites repeat. Then earn a place there. Ask happy customers for Google reviews. Apply to the industry directories that keep appearing. Pitch the local paper and trade sites that publish “best of” lists. Do not buy reviews or paid placements that pretend to be rankings.

Cost and time. Often free, plus your time. Expect weeks to months.

Cause 2: No page answers the exact question

What I saw. ChatGPT asked buyers “tank or tankless?” and “residential or commercial?” Those are the questions a good page answers. In my Kigali audit, 5 of 11 AI answers cited a business’s own site, mostly for specific questions. Broad questions cited directories instead. In this test, 9 of the 12 plumbing and HVAC sites I could read linked from the homepage to a page about the exact job. That was common among named businesses, so it did not separate winners from the rest. Still, the three that lacked such a link were named anyway, which suggests other signals, such as reviews, carried them (see cause 1).

The fix. Write one page per job a buyer asks about. Open with a two-sentence answer. Then cover the follow-up questions: which type, which size, which price range, which area, how fast. The Lens & Co case shows what plain facts can do. Delivery windows, a price floor and lens measurements were repeated closely in AI answers.

Cost and time. $0 if you write it. $80 per article if you hire me. Allow about 2 weeks before you test.

Cause 3: AI cannot tell which business you are

What I saw. The same business showed up under two names. ChatGPT listed “Tech Ridge Dental” at 12400 N. IH-35. Perplexity listed “MA Dentistry, PLLC” at 12400 N IH-35. The practice’s own site says Tech Ridge Dental, at 12400 N. Interstate 35, Ste A131. It looks like one practice with two names. Likewise, ChatGPT named Integris and Perplexity named TechMD. The TechMD website now redirects to Integris, so I treated them as one company. I counted each pair once. One more case: I could not match one of ChatGPT’s four roofing picks to any website with an Austin phone number. Location is blurry too. Google’s AI Overview led its Austin CPA answer with a firm listed in Katy, Texas, and Perplexity named the same firm.

The fix. Use one business name, one address and one phone number on your site, your Google Business Profile and every directory. If you rebranded, say so on your site and update old listings. Google also advises checking that Business Profile information is up to date. Add LocalBusiness structured data that matches what visitors see.

Cost and time. Free if you do it. $180 if I do the Business Profile and directory clean-up. Days to weeks, because directories update slowly.

Cause 4: Your pages say nothing quotable

What I saw. Buyers care about price, speed and proof, and ChatGPT’s roofing form lists those priorities. Yet only 10 of the 32 named local sites I could read (31%) had a price or cost link on the homepage. Among the 31 national sites, 21 did (68%).

The fix. Say what you do, who it is for, what it costs, how long it takes and why you can be trusted. Add a price range or a starting price. Name your people and your licenses. Show real photos, real numbers and real results. Google’s guidance is to create “helpful, reliable, people-first content.” A page of general claims gives an engine nothing to repeat.

Cost and time. Your time, or $80 per article. Plan on 2 to 4 weeks for a first round.

Cause 5: Google or Bing has not indexed you

What I saw. None of the 63 named sites carried a noindex tag. That is no surprise, because established businesses rarely have this problem. New sites and relaunches do. Google says a page “must be indexed and eligible to be shown in Google Search with a snippet,” and it adds that indexing is not guaranteed.

The fix. Verify your site in Google Search Console. Submit your sitemap. Inspect your key pages and request indexing. Check that no page carries a noindex tag. Then do the same in Bing Webmaster Tools.

Cost and time. Free. Days to weeks.

Cause 6: Basic page signals are missing

What I saw. The 63 named sites had the basics. In the raw HTML, 90% had a meta description and 94% had at least one H1. Another 83% carried business schema such as Organization or LocalBusiness. By contrast, my Kigali audit found 9 of 20 sites with no meta description and 8 of 20 with no H1. So basics look like table stakes in the markets I tested. Fix them, but do not expect them to be your main problem.

One caution on llms.txt. Twenty-eight of the 63 sites had one, and many look generated by an SEO plugin. I cannot show that it helped. Google says you do not need “new machine readable files, AI text files, or markup” to appear in its AI features. I cover this in my llms.txt guide.

The fix. Give every key page a unique title, a meta description, one clear H1 and matching schema. Keep structured data identical to the visible text.

Cost and time. Free with an SEO plugin. $220 for my schema pack. One to two days.

Cause 7: Crawlers are blocked

What I saw. This is the cause most guides list first. In my sample it explained nothing. None of the 63 named sites blocked an AI search bot in robots.txt, and none blocked a training bot. The same was true of the 20 Kigali sites. Even so, the risk is real. OpenAI says that “sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers.” Perplexity recommends allowing PerplexityBot. And since July 1, 2025, Cloudflare blocks AI crawlers by default unless the owner allows them.

One more note. Fourteen of the 77 sites refused or failed my plain request: ten returned 403, two returned 202 with no page, one returned 404 and one would not connect. That does not prove they block AI crawlers, because firewalls treat bots differently. Still, it is why you should test your own site.

The fix. Open yourdomain.com/robots.txt. Look for “Disallow: /” under any bot. Then check your firewall or CDN bot settings. Allow OAI-SearchBot, PerplexityBot and Claude-SearchBot. Training bots such as GPTBot are a separate decision. My robots.txt guide has examples.

Cost and time. Free. OpenAI says it can take about 24 hours for its systems to adjust after a robots.txt update.

Which cause is yours? A decision table

Match what you see to the likely cause. Then start with the fix in the last column.

What it costs to fix

Most fixes cost time, not money. The table shows what you can do yourself and what I charge. These are the prices on my pricing page on November 5, 2026.

For a bigger job, my Standard build costs $850 and covers up to 30 pages, five answer-first pages and an internal linking plan. My Growth retainer costs $650 a month. For comparison, one vendor price guide quotes $2,000 to $10,000 a month for mid-market AEO and GEO retainers, and $500 to $2,500 a month for productized packages. That same guide gives different ranges in different sections, and agency blog figures are not contracts. Treat them as a rough guide, and ask any vendor to list deliverables.

How long it takes

Nobody can promise a date. Here is what the evidence supports.

Because AI answers change run to run, test the same questions every month and log the result. One good answer is not a trend.

The 30-minute test

Run this before you spend a dollar. It shows which cause is yours.

1. Write 10 questions (5 minutes). Use real buyer wording: “Who is the best [service] in [city]?”, “Best [service] for [type of customer],” “How much does [service] cost in [city]?” and “[Service] for [specific problem].”
2. Run each one (10 minutes). Use ChatGPT with web search on, Perplexity and Google. Use a private window where you can, and add the city if you are local.
3. Log the result (5 minutes). Copy the table below into a sheet.
4. Click the sources (5 minutes). Write down which list and review sites repeat. Those are your cause 1 targets.
5. Check the basics (5 minutes). Open your robots.txt file. Inspect your key page in Search Console. Compare your Business Profile with your website. Search your own site for a page about the job.

For a deeper version, see my guide to checking what ChatGPT says about your business and my post on why Google rankings do not match ChatGPT mentions.

Where I come in

I spent six years writing for crypto and finance outlets, where a wrong fact costs readers money. Now I test what AI engines say about brands and publish the data. If you want a second pair of eyes, the free audit scans your site, scores it out of 100, lists your top fixes and checks what ChatGPT, Gemini and Perplexity say about you. I cannot promise a ranking or a citation. I can promise a clear fix list, sourced data and honest reporting.

Run the free AI visibility audit

Contact me, or see my pages on AEO and GEO.

Every answer I recorded

You should not have to take my word for the numbers. Below is the top of every answer I recorded: ChatGPT, Perplexity and Google, for each of the 10 questions. The family dentist screenshots are in full above. Open a question to see its three screens.

Best plumber for a water heater replacement, Austin

Best CPA for a small business, Austin

Best roofing company, Austin

Best HVAC company for AC repair, Austin

Best bookkeeping service, US

Best managed IT provider, US

Best web design agency, US

Best virtual assistant company, US

Best HR consulting firm, US

Perplexity: not run. Its free search limit ran out after nine searches.

How I checked, and the limits

• One snapshot. I ran all searches on November 5, 2026. Answers vary run to run and by user. Google varies by day and location.
• Accounts. ChatGPT was on a free plan, signed in, with web search on. Perplexity was on a free plan, signed in. Account memory shaped the end of some ChatGPT answers: the IT, virtual assistant and web design answers asked whether I was a customer or a competitor researcher. It may also have shaped other details.
• Perplexity gave two answers to the plumber question. I ran it twice because my first read missed names. Cold Is On The Right led the first answer and Radiant led the second. I used the second. I also counted businesses named in the text of Perplexity’s answers, not only in its tables.
• Perplexity stopped at nine. Its free search limit ran out, so I did not run the HR question there. I did not work around the limit.
• Wording. Austin questions named the city. The US-wide questions ended “in the US” on ChatGPT and Perplexity, and had no location words on Google.
• Google. I read the first 8 organic results of each search with US settings and personalization off. “Own page” means the business’s own domain appeared among them. Directories, forums, media lists and review sites did not count.
• Matching names to websites. I matched each named business to a website, then checked the page. Of 82 listings, 64 were verified, 16 were likely matches and 2 had no match. I counted a business once when two listings pointed to the same domain.
• The website audit. I fetched each homepage and robots.txt once, with a plain, honest user agent. I read raw HTML only, so content that JavaScript adds was invisible. Fourteen of 77 sites refused or failed, and I left them out. I did not retry or work around a refusal.
• “First named” means the first business in the answer or the first row of Perplexity’s table.
• Source chips are partial. ChatGPT showed source names on some paragraphs and not others.
• Small samples. These are 10 questions and 63 sites, not a random sample. The audit shows what named sites have. It does not prove what caused a mention. The ranking of causes is my judgment.
• Outside data. I read BrightLocal, Pew, Ahrefs and the Google and OpenAI pages on their own sites. The price guide is a vendor’s blog post. Check the originals before you reuse a number.

Related reading: AI citation vs AI recommendation, how to get cited by ChatGPT and SEO, AEO and GEO explained.