Content Marketing From the Search Side: The Text Google Finds and AI Cites

The spreadsheet arrived on Monday. Twelve posts a month, one for each keyword. “What Is X,” “Five Tips for Y,” “The Complete Guide to Z.” On the last line sat the usual promise: stay consistent, and the blog starts bringing in clients from Google.

Before you approve it, look at what search itself says. On September 14, 2026, I searched Google in Brazil, where we are based, for “marketing de conteúdo.” That is Portuguese for content marketing. The AI Overview opened the page with a definition. Content marketing, it said, is a strategy to create and distribute relevant, valuable, and consistent material. ¹

Those three adjectives are the ones in the Content Marketing Institute’s definition, which has been on the institute’s page since 2015. ^{2 3} None of them says how search decides whether your text has value.

Google, which makes that call in search, has published its own questions. This piece gathers what Google documents about content made for people, about AI in production, and about content at scale. At the end, your spreadsheet goes through a five-column test.

The term, and the part of it that belongs here

The most repeated definition is the Content Marketing Institute’s. It calls content marketing “a strategic marketing approach.” ² The approach focuses on creating and distributing valuable, relevant, and consistent content. The aim is to attract and retain a clearly defined audience. And, in the institute’s words, “ultimately, to drive profitable customer action.” ²

The institute’s page carries no date. Copies in the Internet Archive tell the story. In November 2012, the definition called the craft “a marketing technique” and spoke of attracting, acquiring, and engaging. ³ Today’s wording shows up in the February 2015 copy. ³

On the same page, the institute lists where good content does its work. Email marketing, social media marketing, SEO, PR, and PPC are on that list. ² Content marketing is bigger than search, and the institute itself shows it.

This site is about search. So this piece stays with one slice of the field: the content someone finds when they ask Google or an AI. Email and social media stay out. The places where search happens today, each with its own rules, are in the piece on search marketing.

What the first page says, and what it leaves out

I read the top ten results for “marketing de conteúdo” in Semrush’s Brazil database, plus the live screen on Google in Brazil. ^{1 4} Together, the two readings gave thirteen addresses. I read eleven pages. The video counted only by its title, and one page would not load. ⁴

All of them define the term and explain what it does for a business. The Portuguese-language Wikipedia quotes the institute’s definition. ⁴ One of the texts talks about AI answers and offers its own tool to measure presence in them. ⁴

Three things did not appear on any of the eleven pages. None cites Google’s page on helpful content. None mentions the policy against scaled content abuse. And none says what Google wrote about using AI to produce text. ⁴

These are three documents from the owner of search. Your spreadsheet goes through all of them, whether anyone reads them or not.

The questions Google published

Google has a page called “Creating helpful, reliable, people-first content.” ⁵ Its first sentence says what the ranking systems are for. They are designed to prioritize helpful, reliable information created to benefit people. ⁵ The same sentence rules out content made to manipulate rankings. ⁵

Next come the self-assessment questions. They let site owners look at their own text with an outsider’s eye. Five of them hit the posts spreadsheet head on. ⁵

• Does the content provide original information, reporting, research, or analysis?
• Does it offer insightful analysis or interesting information that goes beyond the obvious?
• If it draws on other sources, does it add real value and originality instead of copying them?
• Would you expect to see it in, or cited by, a printed magazine, encyclopedia, or book?
• Does it provide substantial value compared with other pages in the results?

The page suggests one more step, and it costs little. Ask people you trust, with no ties to your site, for an honest assessment. ⁵

Another block asks whether the text is made for people. Do you have an audience that would find the content useful if they came to you directly? Does the text show first-hand expertise, from someone who actually used the product or visited the place? After reading, will someone have learned enough to reach their goal? ⁵

The question about the site’s focus sits in that block too. What it has to do with the size of your territory is in the piece on topical authority.

The warning signs that look like an editorial plan

The same page carries a second list, and Google tells you how to read it. Answering yes to some of these questions is a warning sign to rethink how you create content. ⁵ Read the list with the spreadsheet next to you.

Is the content made mainly to attract visits from search engines? Are you producing lots of content on many topics, hoping some of it performs? Are you using extensive automation to produce content on many topics? ⁵

Are you mainly summarizing what others say, without adding much? Are you writing about something only because it is trending? Did you move into a niche with no real expertise, because you thought it would bring traffic? ⁵

Two questions take apart common blogging playbooks. One is about word count. Google asks whether you write to a set word count because you heard it has a favorite. And it answers, in parentheses: “No, we don’t.” ⁵

The other is about dates. Changing a page’s date to make it look fresh, without really changing the content, is on the warning list. So is publishing or removing lots of content to make the site seem fresh. On that one, Google answers that it won’t work. ⁵

Go back to the twelve-post spreadsheet. In my reading, a content plan built only on keyword volume answers yes to the first question on the list.

So isn’t SEO just writing for search engines? The page takes up that very doubt. It describes SEO as the things that help search engines discover and understand content. And it calls SEO a helpful activity when applied to people-first content. ⁵ Choosing the right question for each page is the subject of the piece on search intent.

What AI asks for, by the same standard

Picture the counter at a hardware store. You ask how to hang a shelf on drywall. The first clerk reads the back of the wall anchor package out loud. The second asks how much weight will sit on the shelf, because he has seen shelves come down.

Both of them answered. Only one told you something you could not have read on the package yourself.

Google has a name for the first clerk’s kind of text. It is in Google’s official guide to its generative AI features in Search, such as AI Overviews and AI Mode. ⁶ The guide describes this content as often based on common knowledge, which could come from anyone. ⁶ This is called commodity content, in Google’s own term.

The guide’s example is a pair of titles. “7 Tips for First-Time Homebuyers” is commodity content. “Why We Waived the Inspection & Saved Money: A Look Inside the Sewer Line” falls outside it. ⁶ The second brings the view of someone who went through it, beyond common knowledge. ⁶

Right after that, the guide asks for a point of view of your own. It gives the reason: AI systems look at a variety of sources, so a viewpoint that stands out helps. ⁶ Google tells you not to recycle what others have already said. Nor what a generative AI model could easily produce. ⁶

Notice also what the guide lets you skip. It says you don’t need to write in a special way for generative AI search. ⁶ And that SEO best practices still apply, because these features are rooted in the core ranking and quality systems of Search. ⁶

The path a text takes to become a citation is in the piece on how to show up in AI answers. Here, the subject is what goes into the spreadsheet.

AI in production: what Google accepts and what it calls spam

On February 8, 2023, the official Search blog published Google’s guidance on AI-generated content. Danny Sullivan and Chris Nelson signed it. ⁷ The post opens with a memory.

About ten years earlier, there were concerns about a rise in mass-produced content written by people. No one would have thought it reasonable for Google to ban all human-generated content. It made more sense to improve the systems to reward quality. ⁷

The rule that comes out of this has two halves. Appropriate use of AI or automation is not against Google’s guidelines. ⁷ Using automation, AI included, mainly to manipulate rankings violates its spam policies. ⁷

One line from the FAQ settles half the meetings: “Using AI doesn’t give content any special gains. It’s just content.” ⁷ If it is useful, original, and satisfies aspects of E-E-A-T, it might do well in Search. If it doesn’t, it might not. ⁷

The next question is more direct. Should you use AI to generate content? If it helps you produce useful, original text, it might be worth considering. If you see it as a cheap, easy way to game rankings, Google’s answer is no. ⁷

The policy that took the tool out of the equation

On March 5, 2024, Google announced three new spam policies. One of them deals with scaled content abuse. ⁸

Google’s spam policies page gives the definition. It happens when many pages are generated mainly to manipulate rankings, without helping users. Typically, it means lots of unoriginal content with little to no value, however it was made. ⁹

The 2024 announcement explains that last part. The policy applies to content made by automation, by people, or by a mix of both. ⁸ Under the rule, who did the typing weighs nothing. What weighs is the purpose, and the value that reaches the reader.

Its list of examples reads like the description of a post factory. Using generative AI to create many pages without adding value. Scraping search results and dressing them up with synonym swaps or machine translation. Stitching together passages from several pages without adding anything. ⁹

The AI guide adds one more case. Creating a page for every variation of a query, mainly to manipulate rankings or AI answers, also violates this policy. ⁶

As for the cost, the announcement spells it out. Sites that violate the spam policies may rank lower in results, or not appear at all. ⁸

Where the tool helps

Google’s guidance on generative AI content on your website says where the tool helps. It can be useful for researching a topic and for adding structure to original content. ¹⁰ The risk starts when it generates many pages without adding value. ¹⁰

The page also asks for accuracy in what appears in search results. Titles, meta descriptions, structured data, and image alt text all count. ¹⁰ And it suggests telling readers how the content was made, in a way that makes sense to them. ¹⁰

On the helpful content page, that idea becomes a question, the “How” question. When readers might wonder how something was created, an AI or automation disclosure is useful. ^{5 7}

AI also cites machine-written text

A recent audit measured the other side, the side of the answer. Mowafak Allaham and Nicholas Diakopoulos, of Northwestern University, published the study in May 2026. ¹¹

They ran 712 real user queries through ChatGPT, Copilot, Gemini, and Perplexity. The queries were in English, on politics, health, and the environment. ¹¹ Then they ran the text of the cited sources through an AI-text detector called Pangram. ¹¹

About 16% of the cited sources were flagged as AI-generated text. ¹¹ The authors treat the number as a floor, because they chose the detector that lets more machine text slip through. ¹¹

The limits come with the number. It is a preprint, with no published peer review. It relies on one detector, on queries in English, and on a United States scope. ¹¹ Other languages were left out of the count.

The reading I take from the study is my own. Being cited by AI does not prove a text is good, because the machine also cites machine text. What sets yours apart is what can be checked: who said it, and where the data came from.

Who writes it, and why that can be checked

Picture two gluten-free cake recipes, identical on paper. One came on an unsigned flyer. The other is on a registered dietitian’s site, with her name and license number. The second one you can check in two minutes. The first you can only take on faith. ¹²

The helpful content page runs that check in three questions. Is the content written or reviewed by an expert or enthusiast who demonstrably knows the topic well? ⁵ Would someone who researched the site come away thinking it is well trusted or widely recognized on its topic? ⁵ Does the content have any easily verified factual errors? ⁵

The word that carries the weight is demonstrably. Knowing the topic is half the job. The other half is someone being able to check it.

In its 2023 FAQ, Google treats the byline the same way. Accurate author bylines make sense where readers might think, “Who wrote this?” ⁷ And giving AI an author byline is probably not the best way to disclose that it took part. ⁷

In my book SEO de Entidade, published in Portuguese, I gave a name to the passage that lets itself be checked. When a passage carries inside it where the information came from, I call it an attributable passage. ¹² The number, the year, and the name of whoever measured it travel in the same sentence. ¹²

The rule holds for the data and for the byline. The path from the byline to the author page, and on to sources outside the site, is in the piece on E-E-A-T. That piece also covers what Google documents about marking up the author in the page’s code.

In my reading, this is where content marketing meets entities. A name the machine can link to a real person, with records outside the site, backs up what the text claims. That recognition is the subject of our semantic SEO section.

The spreadsheet test

The first snapshot comes out today, from the posts spreadsheet or from the last ten posts on the blog. Take a sheet of paper and draw five columns.

1. Question: which customer question does this text answer, in the customer’s own words?
2. Only ours: what does it have that a competitor’s would not, such as data, a case, a photo, or experience?
3. Byline: who signs it, and where does the name lead?
4. How it was made: if AI was used, what did the team add, and would the reader know?
5. Magazine: would you expect to see this text in a printed magazine or book? ⁵

Now read the sheet. A row with column 2 blank is commodity content, in Google’s term. ⁶ It goes into the queue to be rewritten with what only you know, or it leaves the spreadsheet.

A row that exists only because the keyword had volume needs a customer question in column 1. Without one, the text has no reason to be written yet.

A word on method. The sheet shows you where to look, and no sheet measures ranking. Google writes that no third-party tool has access to its internal ranking or AI systems. ⁶

The question you hand back

From now on, when the next twelve-post spreadsheet arrives, you have a question to hand back. What does each text bring that is not already in the top ten results?

Anyone who answers with a word count or a posting frequency is answering a different question. Google wrote that it has no preferred word count. ⁵

Print the list of the last ten posts on the blog today. Fill in column 2 before the next content planning meeting, and bring the sheet to the table.