Interviews, Professional Services & Technology

Aarón Gómez answers: how a Panamanian small business can use AI for content without losing credibility

Daniela Bohórquez, redactora de Radar PanamáDaniela Bohórquez711 Mins Read
Aarón Gómez answers: how a Panamanian small business can use AI for content without losing credibility

In short: Aarón Gómez, a specialist in web positioning and AI-generated answers (SEO, GEO and AEO), argues that a small business can use AI for the first draft of its content, but that a person with judgment must verify and sign it. His rule: AI speeds up the form, it does not replace judgment or the source. He proposes checking every fact against its source in five steps, anonymizing any customer information before pasting it into a tool, and setting aside two hours a week to answer customers’ real questions in writing. According to its documentation, Google evaluates the quality of content and not how it is produced, although it asks for AI output to be reviewed manually.
Contents of this interview
  1. Who is answering and why it matters
  2. Which tasks yes and which not yet
  3. How each fact is verified, step by step
  4. How to spot unreviewed AI text
  5. Which tools to start with
  6. Customer data and AI
  7. A two-hour weekly routine
  8. What the data says
  9. Practical summary
  10. Frequently asked questions
  11. Sources

Who is answering and why it matters

Aarón Gómez is a specialist in web positioning and in artificial intelligence answers (SEO, GEO and AEO), with 8 years of experience. He holds a business administration degree with a focus on accounting systems, works with projects in Latin America and North America, and collaborates with Radar Panamá. His professional profile is on LinkedIn.

Daniela Bohórquez, a Radar Panamá reporter, put six questions to him in writing and he answered on November 6. The answers are published verbatim, as he sent them. The topic matters to any Panamanian small business owner who wants to use AI for content and marketing without losing credibility or exposing customer data. This interview complements the one with Néstor Romero on AI and jobs, which deals with positions and universities, while this one deals with the content and marketing of a small company: Néstor Romero answers: which jobs AI changes first in Panama and what universities are missing.

Answers translated from Spanish by Radar.

Which tasks yes and which not yet

For which content and marketing tasks is it already worth using AI in a small business, and for which not yet?

«It is worth it for: first drafts, headline variations, summaries of reviews, ideas for frequently asked questions, translation of material that has already been validated. It is not worth it for: verifiable facts (prices, deadlines, figures), testimonials, customer cases, and any content that requires direct experience or legal responsibility. The practical rule: AI produces the first draft, and a person with judgment verifies it and signs it. If the final text has nothing that only you can know, it is probably not worth publishing. AI speeds up the form; it does not replace judgment or the source.»
Aarón Gómez, SEO, GEO and AEO specialist

The split he proposes matches Google’s public position. According to the Google Search Central blog, the search engine focuses on the quality of content and not on how it is produced, and using AI to generate content with the main purpose of manipulating ranking violates its spam policies. The guidance on content generated with AI adds that it is best to focus on accuracy, quality and relevance, and to manually review what the tool generates.

How each fact is verified, step by step

You say you check every fact against its source. What does that process look like, step by step?

«Step one: I identify every claim a reader could question: figures, dates, names, comparisons. Step two: for each one, I look for the primary source or the closest one possible: official documentation, a study, client data. Step three: if I find no source, the claim is removed or rewritten as an explicit opinion (“in my experience…”). Step four: I note the source next to the fact in the draft, even if it is not published, so I can answer if someone asks. Step five: a second read looking specifically for numbers and names. It is not glamorous, but it is the only way to publish without risking credibility.»
Aarón Gómez

1. Identify

Every claim a reader could question: figures, dates, names, comparisons.

2. Find the source

The primary one or the closest possible: official documentation, a study or client data.

3. Remove or soften

With no source, the claim is dropped or rewritten as an explicit opinion.

4. Note the source

Next to the fact in the draft, even if it is not published.

5. Reread numbers and names

A second read devoted only to them.

The method fits Google’s instruction to manually review what AI produces. There is also a measured reason not to trust a first draft: the Stanford HAI AI Index 2026 reports that, in the AA-Omniscience benchmark, the hallucination rate of 26 models ranges from 22% to 94%. Radar cites it only as an example that models make mistakes, not as a general error rate for everyday AI use.

How to spot unreviewed AI text

How can you tell a text was written by AI without review, and what does it cost a brand?

«You can tell by a combination of things: sentences that sound good but say nothing concrete (“in today’s fast-paced world”), an overly symmetrical structure, no specific examples, and zero verifiable facts. Also by the lack of a voice: everything in a neutral third person, with no opinion and no experience. What the brand loses is credibility. A reader who spots generic text assumes the company has nothing of its own to say. And if the text contains a false fact, the loss is greater: broken trust is hard to recover. Unreviewed AI is like a bank brochure: correct, empty, forgettable.»
Aarón Gómez

The risk is not only one of image. Google’s documentation on AI-generated content warns that producing many pages with no value for the reader can violate its scaled content abuse policy. For a small business, the way out Gómez describes is the opposite: fewer pages, each with something of its own to say.

Which tools to start with

Which tools would a small Panamanian company use to get started, and how does it choose among them?

«I would start with what it already has: ChatGPT or Gemini in the free version for drafts and brainstorming, Google Docs for writing and commenting, and Google Search Console to see what people are searching for. To choose between tools, the criterion is not the most popular one, but these: does it save me time on a task I already do? Can I verify what it produces? Is my data protected? For a small business, the learning curve matters more than power. A tool that is used every day is worth more than one that was tried once.»
Aarón Gómez

The three criteria he mentions (time saved on a task that already exists, the ability to verify the result, and data protection) work for comparing any tool, including those that appear later. The local market context helps size the audience this content reaches: according to DataReportal, Panama ended 2025 with 3.60 million internet users, 78.4% of its 4.58 million inhabitants.

Customer data and AI

What risks does customer data face when AI is used in marketing?

«The main risk: pasting customer information (names, emails, phone numbers, details of conversations) into public AI tools without knowing how it is stored or whether it is used for training. That can violate privacy expectations and, in some cases, regulations. The safe practice: anonymize before using AI (no names, no contact details), do not upload entire databases, and prefer tools with clear policies of not training on user data. If the data identifies a person, it should not go into a prompt without consent and without a real need.»
Aarón Gómez

Radar note: Panama has a personal data protection law, Law 81 of 2019. Anyone who handles customer data should review it with their legal adviser; this article does not go into its content. Nor does it claim anything about how each tool treats the data it receives: as the interviewee says, the data policy of each one should be checked before using it.

A two-hour weekly routine

What two-hour weekly routine would you recommend to a business owner who wants to use AI?

«One hour: review the real questions of the week (WhatsApp, calls, emails), pick three, and write with AI a draft answer for each one. Thirty minutes: verify each draft against the reality of the business (prices, deadlines, availability) and correct it. Thirty minutes: publish one of those answers on the website or the Google profile, and note the other two for the following week. Two hours a week, with human verification, produce more useful content than five hours of mass generation without review.»
Aarón Gómez

The routine starts from the questions customers already ask, so the content answers a real need and not an invented list of topics. The closing comparison with mass generation rests on the same Google warning about producing valueless pages at scale.

What the data says

Data pointWhat it saysSource
Quality, not methodGoogle focuses on the quality of content, not on how it is producedGoogle Search Central, February 2023
AI used to rankGenerating content with AI with the main purpose of manipulating ranking violates spam policiesGoogle Search Central, February 2023
Manual reviewFocus on accuracy, quality and relevance, and manually review what AI generates; producing many pages with no value can violate the scaled abuse policyGoogle Search Central, generative AI content
Model errorsIn the AA-Omniscience benchmark, the hallucination rate of 26 models ranges from 22% to 94%Stanford HAI, AI Index 2026
Online audience3.60 million internet users in Panama at the end of 2025 (78.4% of the population)DataReportal, Digital 2026 Panama

Caveat: the hallucination figure is an example from one specific benchmark, not a general error rate. Radar did not find a verified figure on how many small businesses in Panama or Latin America use AI in their marketing, so none is offered.

Practical summary

  1. Ask AI only for the first draft. Alternative headlines, review summaries, ideas for frequently asked questions or translations of material already validated.
  2. Leave out what carries responsibility. Prices, deadlines, figures, testimonials and customer cases are written and verified by a person.
  3. Verify every claim against its source. If there is no source, it is removed or presented as an opinion.
  4. Anonymize before pasting. No names or contact details, no entire databases.
  5. Set aside two hours a week. Three real customer questions, one draft per question, verification and one published answer.
Key fact: according to Aarón Gómez, AI produces the first draft and a person with judgment verifies it and signs it; Google, for its part, says it evaluates the quality of content and not how it is produced.

More interviews from Radar Panamá: Aarón Gómez answers: why your website brings no clients and what to do first · Aarón Gómez answers: how to check if AI mentions your business · Aarón Gómez answers: how to get Google and AI to use your text as the answer · Aarón Gómez answers: where a Panamanian small business with a tight budget should start digital marketing · Néstor Romero answers: why AI recommends one Panamanian business and not another · Néstor Romero answers: which jobs AI changes first in Panama and what universities are missing

Frequently asked questions

Which marketing tasks is it worth using AI for in a small business?

First drafts, headline variations, review summaries, ideas for frequently asked questions and translation of material already validated, according to Aarón Gómez. It is not worth it for prices, deadlines, figures, testimonials, customer cases or content that requires direct experience or legal responsibility.

Does Google penalize content made with AI?

According to Google, it focuses on the quality of content and not on how it is produced. It does treat as spam the use of AI with the main purpose of manipulating ranking, and its documentation warns that producing many pages with no value can violate the scaled content abuse policy.

How is an AI-made text reviewed before publishing?

With Gómez’s five-step process: identify every claim a reader could question, find its primary source, remove or rewrite what has none, note the source next to the fact and do a second read focused on numbers and names. Google also asks for AI output to be reviewed manually.

What customer data should not be pasted into an AI tool?

Names, emails, phone numbers and details of conversations, as well as entire databases. Gómez recommends anonymizing before using AI and not putting data that identifies a person into a prompt without their consent and without a real need.

How much weekly time does a small business need to start with AI?

Two hours, according to Gómez’s routine: one hour to pick three real customer questions and draft answers with AI, half an hour to verify and correct them, and half an hour to publish one answer and note the other two for the following week.

Sources

Radar answers

Ask about this article

Type your question and we will show you the answer inside this article or in other Radar Panamá publications.

Was this article useful?
By Daniela Bohórquez

Daniela Bohórquez is a writer at Radar Panamá. she writes about travel, tourism and business, with special attention to reviews of destinations, hotels and services. She lives between Panama, Italy and New York, which gives her a comparative view of how business is done in each market.