How to Check an AI Blog Post Before You Publish It (a 12-Point Checklist)

A made-up study looks just as tidy on the page as a real one. So does last year's price. To check an AI-written post, trace every fact, link and product detail to a real source, make sure it adds something the pages that already rank do not, then fix the title, meta description, schema and alt text and read it once as the reader. Here are the 12 checks, with what Google's own guidance says.

How do I check an AI-written blog post before it goes live?

Run the 12 checks below, in order. If you are short on time, do checks 1 to 4: facts, links, invented sources and product details. Those are the errors that cost trust, and in a store they can cost sales. If the post needs to rank, add checks 5 to 7, because a correct post that says what every other page says gives Google no reason to pick it. Checks 8 to 12 polish the page. On a post you plan to keep live for years, skip none of them.

What does Google say about AI content?

Google does not ban AI content. It judges the page, not the tool that wrote it.

In its 2023 post on AI-generated content, Google wrote: "Appropriate use of AI or automation is not against our guidelines." The same post is blunt about rankings: "Using AI doesn't give content any special gains. It's just content." A useful, original page might do well in Search. One that isn't might not.

For a store owner, the rule that matters most is in Google's guidance on using generative AI content, last updated in October 2026: "It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing." The page gives the reason too: "generative models don't retrieve facts, but predict a likely sequence of words based on their training data." And the review doesn't stop at the text you see. It covers the page title and "meta description elements, structured data, and alternate texts for images".

Where Google does step in is bulk content with nothing added. Its spam policy on scaled content abuse describes the problem as "when many pages are generated for the primary purpose of manipulating search rankings and not helping users." One of its examples is "Using generative AI tools or other similar tools to generate many pages without adding value for users".

For the history behind these rules, see our breakdown of what Google's AI content policy says.

What are the 12 checks?

1. Trace every fact and number to a source you opened.
2. Click every link.
3. Remove invented studies, quotes, reviews and case studies.
4. Match product names, prices and specs to your real catalog.
5. Check that the post answers the search it targets.
6. Compare it with the pages that already rank.
7. Add first-hand detail only you have.
8. Read it out loud and cut stock AI phrases.
9. Add internal links with clear anchor text.
10. Check the title, meta description and schema.
11. Check every image and its alt text.
12. Do a final read as the reader.

1. Trace every fact and number to a source you opened

Highlight every number, date, name and claim of fact. For each one, find the page it came from and open it. Does the page say exactly that? Is it the original source, or a blog quoting another blog? Is it still current?

If you cannot find a source, cut the claim or rewrite it without the number. A plain true sentence beats a precise false one. Google's own self-check asks: "Does the content have any easily-verified factual errors?"

Take extra care with health, money and safety topics. There, a wrong number can hurt the person who trusts it.

2. Click every link

Open each link in the post, internal and external. Check three things: the page loads, it is the page the sentence describes, and it still says what the post claims. A model that predicts likely words can also predict a likely-looking web address that does not exist. And links die. A page that worked when the draft was written may return an error today.

3. Remove invented studies, quotes, reviews and case studies

A draft may say "a recent study found" with no study behind it. It may quote an expert who never said those words, or describe a customer who does not exist. Search for the exact study title and the exact quote. If you cannot find them, delete them.

In the US, invented reviews also carry legal risk. In August 2024 the Federal Trade Commission announced a final rule banning fake reviews and testimonials. It covers reviews that claim to come from people who do not exist, "such as AI-generated fake reviews", and it bans businesses from creating or selling them.

The same goes for authors. Google's helpful content guidance calls fake creator profiles, "such as by using AI-generated headshots, made-up names, or false credentials", a form of deception. Our guide to author bios for an AI-assisted store blog covers how to do bylines honestly.

4. Match product names, prices and specs to your real catalog

A post can give a product a slightly wrong name, an old price, a size you do not sell or a feature it does not have. Open your product admin next to the post and check every product mention: name, price, sizes, colors, materials and the link. If a price changes often, leave it out of the post and let the product page carry it.

Check your policies too. The post should not promise shipping times, returns or warranties that your store does not offer.

5. Check that the post answers the search it targets

Search the post's main keyword yourself and look at what comes up. Guides, product pages, comparison lists or videos? That tells you what the searcher wants. If people want step-by-step help and your post is an opinion piece, it will struggle. Then check that the answer sits near the top, not after five paragraphs of warm-up.

6. Compare it with the pages that already rank

Open the top few results for the keyword and read them. Then ask: what does my post add that these do not? If the honest answer is "nothing", you have what Google calls commodity content. Its guide to generative AI search uses "7 Tips for First-Time Homebuyers" as the example of content that "typically adds little unique insight for readers", and says: "Don't just recycle what others on the internet have already said, or could easily be produced by a generative AI model."

Also make sure the post does not follow one source too closely. Google asks whether content that draws on other sources avoids "simply copying or rewriting those sources". Paste two or three distinctive sentences into Google inside quote marks and see what comes back.

7. Add first-hand detail only you have

This is the fix for check 6. A model cannot know what happened in your shop last month. You can. Add one or two things per post that come from your own work: a photo you took, a test you ran, a question customers ask every week, a mistake you made and fixed. Google's checklist asks whether content shows "first-hand expertise and a depth of knowledge (for example, expertise that comes from having actually used a product or service, or visiting a place)". Our ten-minute edit for AI drafts shows where to add those lines.

8. Read it out loud and cut stock AI phrases

Read the post out loud, or let your phone read it to you. Your ears catch what your eyes skip, like five sentences in a row of the same length, or a paragraph that repeats the one before. Cut phrases readers now tie to machine writing, like "in today's fast-paced world", "it's important to note that", "plays a vital role" and "whether you're a beginner or a pro". Cut strings of three adjectives and openings that announce what the section is about to say.

While you read, fix spelling and break up long paragraphs. One idea per paragraph is a good rule.

9. Add internal links with clear anchor text

Link the post to two or three related pages on your site, such as a product, a category or a related guide. Then link to the new post from an older one, so readers and search engines can find it. Google's link best practices say: "Every page you care about should have a link from at least one other page on your site." The anchor text, the clickable words, should be "descriptive, reasonably concise, and relevant to the page that it's on and to the page it links to." Swap "click here" for words that say what the reader gets. Here is how to plan those links as topic clusters for a store blog.

10. Check the title, meta description and schema

Readers rarely see these on the page, so they are easy to forget. Google's fact-check rule covers them anyway.

• Title: it should match the post and not oversell it. Google asks whether the title avoids "exaggerating or being shocking in nature".
• Meta description: this is the summary that can show under your title in search results. Write a short, true one with the point first. Google says the snippet "is truncated in Google Search results as needed, typically to fit the device width."
• Schema: this is the hidden markup that tells Google the page has an FAQ, a how-to or an article. If the post has it, check that the questions and steps in the markup are the ones on the page. Then run the URL through Google's Rich Results Test.

For more on the first two, see our rules for meta titles and descriptions on AI posts.

11. Check every image and its alt text

Look hard at each image. AI-made images can get details wrong, like garbled text on a label or a product that looks nothing like yours. Never use one to show a real product unless it matches what the customer will receive.

Then check the alt text, the short description of the image that search engines and screen readers use. Google's image guidelines say to "focus on creating useful, information-rich content that uses keywords appropriately and is in context of the content of the page," and warn that keyword-stuffed alt text "may cause your site to be seen as spam." If you use AI images in Google Merchant Center listings, Google's AI content page says the image file must carry a tag that marks it as AI-made: IPTC DigitalSourceType TrainedAlgorithmicMedia metadata. More on writing alt text well: alt text as product SEO.

12. Do a final read as the reader

Last, read the whole post once as the person who searched for it, not as the editor. Do they get the answer early? Would they need to look elsewhere? Google lists that second one as a warning sign: "Does your content leave readers feeling like they need to search again to get better information from other sources?" If the answer is yes, the post is not ready.

Should you tell readers AI helped write the post?

Yes, when a reader would wonder. Google's 2023 post says: "AI or automation disclosures are useful for content where someone might think "How was this created?". Consider adding these when it would be reasonably expected." It also advises against naming AI as the author: "Giving AI an author byline is probably not the best way to follow our recommendation to make clear to readers when AI is part of the content creation process."

In practice, put a real byline on the post (a person or your team). Add a short note on how it was made where readers would expect one.

Can software do the check for you?

Partly. Software is good at the mechanical checks: dead links, missing alt text, title length, schema errors and stock phrases. It cannot tell whether your price changed last week, or whether a quote is real, unless it checks the source. Checks 1, 3, 4 and 7 stay with you.

This blog is written by PostSprout, and parts of this list run inside the product. Every PostSprout article is checked and fixed before it is saved: meta title and description length, the target keyword, FAQ schema, internal links, image alt text and stock AI phrases. On this blog we also run a review step that reads each article before it goes live. It checks accuracy, originality, structure, SEO and voice, fixes what it can, and holds the article as an unpublished draft if a problem is left. In October 2026 it held some of this blog's own articles. One was held on October 27 because its closing paragraph only repeated the rest of the post.

Where should you start?

Print the 12 checks, or keep them open next to your editor, and run them on your next post before you press publish. If you want posts where meta length, alt text, FAQ schema and internal links are already handled, so your time goes on facts and first-hand detail, see how PostSprout works.