GEO content retrofit: 5 steps to get your old articles cited by AI

Why your old articles don’t get cited by AI

Most e-commerce blogs were written for Google: keyword-stuffed, padded to hit a word count, search terms crammed into the title. These articles might rank on Google, but Perplexity and ChatGPT won’t touch them.

The reason is straightforward. When AI engines generate an answer, they need to pull a sentence or two from your article and use it directly. If your first three paragraphs are “setting the scene” without a single sentence that answers a question, the AI skips you and grabs the source that gets to the point.

You don’t need to throw everything out and start over. Most of the time, changing titles, rewriting openings, and adding some structure is enough. Here’s how.

Step 1: Pick which articles are worth retrofitting

Not every article deserves the time. Go through your content library and filter by two things.

First, search volume. Open Google Search Console and check which articles still bring in impressions and clicks. If people are searching for the topic, that article has a better shot at getting cited after you fix it up.

Second, information density. A 2,000-word article that says one thing in ten different ways isn’t worth the effort. Prioritize articles that already contain specific data, step-by-step processes, or product comparisons.

I usually start with 10-15 articles, ranked by expected payoff, and work from the top. If your content library is small, just go through all of them. The next steps don’t take that long per article.

Step 2: Rewrite titles as questions

Traditional SEO titles look like this: “Best Shopify SEO Tools 2026.” That works for Google rankings, but AI engines match content against user questions, not keyword strings.

Nobody types “Best Shopify SEO Tools 2026” into Perplexity. They ask “what SEO tools should I use for my Shopify store” or “how do I do SEO on Shopify.”

The fix is simple. Turn your title into something a person would actually say:

Do the same for H2 and H3 headings. “Tool Overview” becomes “What does each tool actually do.” “Important Notes” becomes “Which mistakes waste your ad budget.”

Not every heading needs to be a question. The point is that each heading should carry information on its own, not just be a label.

Step 3: Rewrite opening paragraphs

Small change, big effect.

AI engines weigh the first 100-150 words heavily when deciding whether to cite your content. If your article opens like this:

In today’s competitive e-commerce landscape, more and more sellers are looking for ways to drive traffic beyond paid ads. Email marketing has long been a popular approach. In this article, we’ll cover everything you need to know about email marketing for your store.

An AI engine can’t do anything with that. No definition, no conclusion, just filler.

Change it to:

Email marketing for e-commerce stores uses automated email sequences to turn visitors into repeat buyers. Unlike paid ads, the marginal cost per email is close to zero. Klaviyo’s data shows an average ROI of 36:1 for email marketing. This article covers how to build a working email sequence from scratch.

In the second version, sentence one is a definition, sentence two has a data point, sentence three tells the reader what the article covers. An AI engine can cite the first sentence to answer “what is e-commerce email marketing” and the second to answer “what’s the ROI of email marketing.”

Go through every article and rewrite the opening. It won’t take long, and it’s the single change most likely to get results.

Step 4: Convert comparison paragraphs to tables

A lot of articles contain comparison information buried in paragraph form. Something like:

Mailchimp’s free plan supports 500 contacts, which works for sellers just starting out. Klaviyo doesn’t have a free plan but integrates better with Shopify. If you’re on a tight budget, Mailchimp makes more sense. If you’re already on Shopify, Klaviyo’s automation features will save you time.

The information is all there, but AI engines struggle to extract a structured comparison from prose. Make it a table:

When someone asks Perplexity “Mailchimp vs Klaviyo,” it will lean toward citing a ready-made comparison table over synthesizing one from scattered paragraphs.

Scan your articles for any “A does this, B does that” structure and consider converting it to a table. Not everything needs a table. The sweet spot is three or more options compared across three or more dimensions.

Step 5: Add data sources and FAQ blocks

Two things left to do.

Cite your sources. Go through each article and find every sentence that contains a number without attribution. “Conversion rates improved by 30%.” Where did that come from? Add the source: “according to Baymard Institute’s 2025 report.” When AI engines choose between multiple sources, attributed data wins over unattributed data.

You don’t need to source every number. Your own test results (“I tested this for 3 months”) and common knowledge don’t need citations. Focus on industry reports and third-party statistics.

Add an FAQ section at the end. Write 3-5 questions readers are most likely to ask as follow-ups. The FAQ generates FAQPage schema, and AI engines read that structured data directly when answering questions.

One thing to watch out for: each answer must stand on its own. Don’t write answers that depend on context from the article body. An AI engine might pull a single Q&A pair without any surrounding text.

You can use ChatGPT to quickly audit what needs fixing:

Review the following article and list:
1. Every sentence with a number but no cited source
2. Every comparison that should be a table
3. Whether the first 150 words contain a citable definition or conclusion

Article content:
[paste your article]

How to check if it worked

Wait a week or two after making changes so AI crawlers can re-index your pages. Then search Perplexity for the topics your retrofitted articles cover and check whether your links show up in the answers.

Some articles will get picked up fast. Others might take a month with no visible change. That’s normal. AI indexing is opaque and there’s no way to speed it up. Just fix what you can and keep producing new content.

One time-saver: after you finish a batch, submit the updated URLs manually in Google Search Console. That’s technically for Google, but AI crawler activity tends to correlate with Google’s index. Costs you nothing to try.