<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Douglas Gouvea Manoel</title>
    <description>The latest articles on DEV Community by Douglas Gouvea Manoel (@douglasgouvea).</description>
    <link>https://dev.to/douglasgouvea</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4237003%2F391826db-dab2-4d35-a38a-f705e8e85b1a.jpg</url>
      <title>DEV Community: Douglas Gouvea Manoel</title>
      <link>https://dev.to/douglasgouvea</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/douglasgouvea"/>
    <language>en</language>
    <item>
      <title>How to check if ChatGPT, Gemini and Perplexity recommend your local business</title>
      <dc:creator>Douglas Gouvea Manoel</dc:creator>
      <pubDate>Tue, 10 Nov 2026 11:56:26 +0000</pubDate>
      <link>https://dev.to/douglasgouvea/how-to-check-if-chatgpt-gemini-and-perplexity-recommend-your-local-business-17a6</link>
      <guid>https://dev.to/douglasgouvea/how-to-check-if-chatgpt-gemini-and-perplexity-recommend-your-local-business-17a6</guid>
      <description>&lt;p&gt;When a customer asks ChatGPT, Gemini or Perplexity for "a good dentist near me", the answer is not a search-results page. It is a short list of names. If your practice is not in that list, you are invisible in that conversation, even if you rank well on Google.&lt;/p&gt;

&lt;p&gt;The instinct is to type your own question once, see whether you appear, and treat that as the answer. That is a weak test. The assistants do not give the same names every time, they do not agree with each other, and a slightly different question can change the result. You need a small, repeatable check, not a single lucky (or unlucky) reply.&lt;/p&gt;

&lt;p&gt;This is a method you can run yourself in about ten minutes, plus a look at what happened when I ran a much larger version of the same test.&lt;/p&gt;

&lt;h2&gt;
  
  
  What "recommended" means here
&lt;/h2&gt;

&lt;p&gt;I am not talking about a paid placement or a badge inside the assistant. I mean something simpler and easier to count: in the answer to a buyer-style question, does the assistant name your business?&lt;/p&gt;

&lt;p&gt;That is a mention, not a ranking. Coming first in one answer and being left out of the next four is not "position 1". The useful number is a mention rate: how often you were named, out of how many answers you collected, on a given day, for a given set of questions.&lt;/p&gt;

&lt;h2&gt;
  
  
  The 10-minute check
&lt;/h2&gt;

&lt;p&gt;Pick the way a real buyer would describe you: the category and the city, not your brand name. "Best dentist in [city] for a checkup" is a buyer question. "Is [your practice name] any good?" is a vanity question. You want the first kind.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;1. Write five to ten questions.&lt;/strong&gt; Stay close to how someone would actually ask for a recommendation. Mix a general one ("best dentist in [city] for a routine checkup") with more specific ones (emergency visit, a neighbourhood, a language, insurance, a specialist). Write them down. You will reuse the same wording later.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;2. Open the three assistants, with web search or grounding turned on if the interface offers it.&lt;/strong&gt; You are measuring the answers people get when the assistant can look things up, not a closed-book chat.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;3. Ask the first question in all three, and write down three things:&lt;/strong&gt; whether your business was named, which other businesses were named, and which websites were cited or linked. Then ask the next question. Do not argue with the assistant or paste your own URL into the prompt. That would measure a different thing.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;4. Repeat each question two or three more times, on the same day.&lt;/strong&gt; Answers move around. One mention in a single reply is not a rate. Three or five replies per question, per assistant, is enough for a first snapshot. If you have five questions and you ask each one three times in three assistants, you will have 45 answers. That is already more informative than asking once.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;5. Compute a mention rate, and keep the notes.&lt;/strong&gt; Mentions divided by answers, overall and by assistant. Also keep a short list of the domains that were cited most often. The rate is the baseline. The cited domains are a hint about where the assistants are looking.&lt;/p&gt;

&lt;p&gt;If you come back a month later, ask the same questions, in the same wording, in the same assistants. That is the only honest way to see whether anything moved.&lt;/p&gt;

&lt;h2&gt;
  
  
  What this looks like with a real set of answers
&lt;/h2&gt;

&lt;p&gt;I ran this method at a larger scale on one mid-size dental practice in a U.S. metro area. I used 25 buyer-style questions, asked five times each, on ChatGPT, Gemini and Perplexity, with web search on. That is 375 answers, collected on a single day. I recorded every practice named and every source cited. This is one practice on one day, not a survey of dentists. The figures below come from that collection and were checked against the analysis file on 25 October 2026.&lt;/p&gt;

&lt;p&gt;The practice was named in 30 of the 375 answers, or 8%. That was sixth among the six practices the assistants named most often in that market. Nobody dominated: the most-named practice appeared in only 11.5% of answers, and four others sat between 8.8% and 11.2%.&lt;/p&gt;

&lt;p&gt;The three assistants also disagreed. The practice that led overall was named in 31% of Gemini answers, 3% of ChatGPT answers and 0% of Perplexity answers. The practice I studied was named in 3% of ChatGPT answers, 9% of Gemini answers and 12% of Perplexity answers. Measuring only one assistant would have told a different story.&lt;/p&gt;

&lt;p&gt;Most of the visibility came from a narrow kind of question. In 18 of the 25 questions the practice was never named in any of the 15 answers (three assistants, five runs). Of its 30 mentions, 14 came from a single hyperlocal question about one neighbourhood. Two of the most general questions — a routine cleaning and checkup, and an affordable dentist without insurance — returned no mentions at all in 15 answers, and a third general question returned a single mention.&lt;/p&gt;

&lt;p&gt;The most-cited source was not a practice website. reddit.com appeared among the cited sources in about 18% of answers, more than any practice's own site. Whether a business is discussed where the assistants look mattered as much as its homepage in this set of answers.&lt;/p&gt;

&lt;p&gt;I am not going to turn those figures into a playbook. They do not say how to get named, and I am not promising that any change will produce a mention. What they do show is why a single prompt is a poor test, and why a mention rate, broken down by assistant and by question, is more useful than a yes-or-no screenshot.&lt;/p&gt;

&lt;h2&gt;
  
  
  If you want a first snapshot without doing it by hand
&lt;/h2&gt;

&lt;p&gt;If you only want a quick sample, there is a free checker that asks a short set of buyer-style questions through Perplexity and reports a mention rate, not a score: &lt;a href="https://check.errandcode.com" rel="noopener noreferrer"&gt;check.errandcode.com&lt;/a&gt;. It is a snapshot, it uses one assistant, and the answers will still vary between runs. It is not a substitute for asking ChatGPT and Gemini yourself, and it is not a ranking.&lt;/p&gt;

&lt;h2&gt;
  
  
  What a fuller report looks like
&lt;/h2&gt;

&lt;p&gt;A complete pass of the method above — three assistants, a longer question list, the sources table and a short list of fixes in priority order — looks like &lt;a href="https://errandcode.com/geo/sample" rel="noopener noreferrer"&gt;this anonymized sample report&lt;/a&gt;. The sample is based on a real collection, with the practice, the competitors and the local domains replaced by labels. It is there so you can see the shape of the output before you decide whether to run the same measurement on your own business.&lt;/p&gt;

&lt;h2&gt;
  
  
  What this does not tell you
&lt;/h2&gt;

&lt;p&gt;It does not tell you how to get recommended, and it does not tell you what will happen next month. AI answers change between runs, by location and over time. A mention rate is a baseline and a way to see whether a later re-test moved, not a guarantee.&lt;/p&gt;

&lt;p&gt;It also does not replace your Google presence. If customers still start with a search box, that work still matters. This check only answers a narrower question: when someone asks an assistant for a recommendation in your category and city, how often is your name in the reply?&lt;/p&gt;

&lt;p&gt;If you run the 10-minute version, keep the questions and the counts. The second run, with the same wording, is the part most people skip, and it is the only part that tells you whether anything changed.&lt;/p&gt;

</description>
      <category>seo</category>
      <category>ai</category>
      <category>chatgpt</category>
      <category>marketing</category>
    </item>
  </channel>
</rss>
