Blog

Why ChatGPT never names your business (and what to do about it)

ChatGPT and other AI answers name businesses they can trust from clear entity signals, useful pages, and third-party mentions, not from ranking alone. Smaller businesses often miss schema, question-led pages, llms.txt and consistent NAP. Start with an AI visibility check, then fix and publish monthly if you want citations without a content team.

Ranking is not the same as a citation

If you run a small business, you have probably felt this shift already. People still Google things, but more of them ask ChatGPT, Gemini, Perplexity, or Google AI Overviews and then act on the names that show up. When your business is not one of those names, you do not lose a ranking. You lose the customer before they ever visit a website.

That is the quiet problem behind being invisible to AI. It is not that your site is broken. It is that answer engines prefer clear entities, structured pages, and sources they already lean on. Ranking on classic search helps discovery for humans who still click. Citation is different. Models pull short, confident answers. They favour pages that lead with the answer, FAQ-shaped sections they can parse, schema they can read, and third-party mentions that confirm who you are.

Most small businesses find out the hard way. A competitor with a thinner product but cleaner answers starts showing up in recommendations. Your traffic looks fine until the questions that used to bring calls start going somewhere else. By the time you notice, months of prompts have already trained the engines on someone else's story.

You do not fix that with one blog post or a slogan about being AI ready. You need a baseline: are you named today across the major engines, for the questions your buyers actually ask? Then you need the work done on a schedule: content that answers those questions from your own facts, schema fixes ranked by impact, and a monthly check that shows whether the citation rate moved.

Checklist: what to fix first

Use this as a practical pass for your own site. Do the cheap, high-signal items before you rewrite everything.

  1. Schema that models can parse. Add Organization or LocalBusiness JSON-LD with your legal name, trading name if different, phone, address or service area, and sameAs links to your real profiles. Add FAQPage on pages that already answer common buyer questions. Product or Service schema helps if you sell clear offers. Keep markup honest. Invented ratings or fake reviews are a fast way to lose trust.
  2. Question-shaped H2s with an answer up top. Rewrite key service and location pages so each section heading matches how a buyer asks, then put a 40 to 60 word direct answer in the first paragraph under that heading. Follow with proof, process, or pricing ranges you are willing to publish. Models extract the lead answer first.
  3. Publish an llms.txt at your domain root. Point crawlers and agents at your key pages, pricing, about, and contact facts in plain language. Keep it aligned with what humans see on the site. If pricing or offers change, update llms.txt in the same release so AI systems do not quote a stale story.
  4. NAP consistency across directories. Match name, address, phone, and website across Google Business Profile, Apple Maps, Bing Places and any industry listings you already use. Inconsistent NAP is a classic reason models hesitate to treat you as one entity.
  5. Earn mentions beyond your own domain. Short answers on industry forums, a clear YouTube explainer, a guest note on a niche association site, and accurate company profiles all help. You do not need a PR agency. You need a few credible places that repeat the same facts about who you are and what you do.
  6. Allow the crawlers you care about. If you block GPTBot, Google-Extended, or similar agents in robots.txt without a reason, you are asking not to be read. Review allow and disallow rules with your developer once, then leave a note so nobody re-blocks them by accident.

How to self-test in ChatGPT, Perplexity, and Google AI Overviews

Pick three to five questions a real buyer would ask. Use the words they use, including city or region if you are local. Examples: best [service] in [city], who does [outcome] for small businesses, [category] near me with [constraint].

Run the same prompt in ChatGPT, then Perplexity, then a Google search that triggers an AI Overview where available. Note whether your trading name appears, which competitor was named instead, and whether the answer cites a page you control.

Repeat weekly for a month and keep a simple log. Date, engine, prompt, named or not, who was cited. That log becomes your baseline. It also stops you from chasing one lucky mention that disappears the next day.

Be honest about limits. A single prompt is not a full citation audit. Models change. Personalisation and location can shift answers. Treat self-tests as a signal, not proof.

The AI visibility check versus a paid plan

Start with the AI visibility check if you only need a score and a short list of gaps. It scores how ready your site is to be cited. It does not ask ChatGPT live yet. That monthly sampling across six engines sits on the paid plans.

Use a paid plan when you already know the gap and need the work done every month. Pemba tracks citations across ChatGPT, Gemini, Google AI Overviews and AI Mode, Perplexity, and Claude, writes content from your own site and publishes it to WordPress or Shopify once you approve it, and sends a short report with the next priorities. Monthly plans are in a 90-day results pilot right now: US$30 for the first 30 days, then the founding price. Apply for the pilot. See pricing and services.

When managed AEO beats DIY

Stay on DIY if you have time to ship schema, rewrite question pages, keep directories clean, and publish monthly yourself. Move to a managed loop when the work keeps slipping behind client work and you want audits, content, and citation tracking on a schedule, month to month, and you can cancel any time. More on the how-to path: how to get your business on ChatGPT.

90-day results pilot

Monthly plans are in a 90-day results pilot right now: US$30 for the first 30 days, then the founding price. Apply for the pilot.

If you only do one thing this week

Run the AI visibility check on your own domain and ask the same buyer question in ChatGPT. If your name is missing, you now have a clear problem statement, not a vague SEO worry. From there you can fix the P1 gaps yourself, or let Pemba run the monthly loop.

Questions people ask

Why doesn't ChatGPT mention my business?

Usually because the model cannot treat you as a clear, trusted entity for that question. Missing schema, buried answers, inconsistent NAP, blocked crawlers, or no third-party mentions all make it easier to name someone else.

Is SEO enough to get cited in AI answers?

SEO helps, but ranking alone is not a citation. Answer engines also need extractable copy, structured data, entity clarity, and corroborating mentions. Classic SEO and AEO work together; one does not replace the other.

What is the AI visibility check actually testing?

Readiness signals: whether your site looks extractable and citation-ready. It is not a live multi-engine citation sample. The monthly service runs that live check every week, on 10 questions, across six engines.

Do I need llms.txt for Google?

llms.txt is mainly for AI crawlers and agents that look for a plain map of your key facts and pages. It does not replace Google SEO. It is still worth publishing so tools that read it get the same story as your site.

How long until AI starts naming us?

SEO improvements often show in weeks. AI citation movement commonly takes longer because models refresh on new web content over time. Pemba's live site notes that AI visibility improvements often take about 2 to 3 months. Your monthly report is there to show trajectory, not a promised share of voice.

What's the difference between a tracker and done-for-you AEO?

A tracker tells you whether ChatGPT and other engines named you. Done-for-you AEO also writes the content, publishes it once you approve it, and lists the schema and on-page fixes. If you only want a dashboard, a cheaper monitor is enough. If you want the work done without agency meetings, choose an implementation plan.

How much does Pemba cost?

Monthly plans are in a 90-day results pilot right now: US$30 for the first 30 days, then the founding price. Apply for the pilot at pemba.ai/contact?type=pilot.

Can Pemba publish to WordPress or Shopify?

Yes. Content is written for you and published to WordPress or Shopify once you approve it by month end. Other platforms get finished content to paste in while native publish expands.

Written by the Pemba team. Last updated 06/10/2026.