AI-SEO Content Audit: 12 Checks for an Existing Blog

If your blog was built for traditional SEO, it is probably invisible to AI. That is not a hypothetical — it is what we see when we audit client sites. Good keyword rankings, reasonable traffic, but zero presence in ChatGPT answers, Perplexity citations, or Google AI Overviews. A thorough seo content audit built for 2026 has to account for both worlds: the classic ten-blue-links result page and the AI answer layer that now sits above it. At Choco Media, we have developed a 12-point checklist we run on every blog before we touch a single word of new content. This post walks through each check, what we look for, and how to fix what we find.

This guide is for marketing managers, founders running their own content, and agencies that have inherited a blog with history. You do not need a developer for most of these checks — a spreadsheet, a browser, and an hour of honest attention will take you far. We have kept the language practical. If something requires a tool, we will name the specific one we use.

By the end you will have a clear picture of which posts are worth saving, which need structural surgery, and which should be quietly retired. You will also understand exactly why certain posts rank on Google but never surface in AI-generated answers — and what to do about it.

Why a Separate AI-SEO Audit Layer Matters Now

Most SEO audit frameworks were built when Google’s result page was the only destination that mattered. That assumption no longer holds. In client work we have found that posts ranking in positions 3–8 on Google are almost never cited by ChatGPT or Perplexity, while posts that answer a specific question directly — even if they rank lower — get picked up by AI systems consistently. The ranking signal and the citation signal are diverging.

The reason is structural. Google’s ranking algorithm rewards authority, backlinks, and topical depth. AI citation algorithms reward clarity, direct answers, and structured data. A post can have all of the first and none of the second. Our audit catches both gaps in one pass.

• AI systems pull from content that answers questions directly, not from content that builds topical authority slowly over time
• Schema markup — particularly FAQPage, Article, and Speakable — is a direct signal to AI crawlers
• Internal linking helps both traditional SEO and AI understanding of your site’s topical structure
• Thin content that pads word count is penalised more heavily in AI systems than in classic search

Check 1: Does Every Post Have a Clear, Answerable Question at Its Core?

This is the first thing we look for, and it is the one that fails most often. A post titled “Social Media in 2026” is not answerable. A post titled “How often should you post on Instagram in 2026?” is. AI systems are answer engines. If your post does not answer a specific question — visibly, in the first 150 words — it will not be cited, regardless of how good the rest of the content is.

Go through your blog index and write down the core question each post answers. If you cannot write it down in one sentence, the post has a structural problem. This exercise alone usually reveals 20–40% of posts that need a headline and opening rewrite before anything else is worth fixing.

• Check that the target keyword appears in the first paragraph, not buried in a subheading deep in the scroll
• The answer to the core question should appear before the first H2 — do not make readers or AI systems hunt for it
• Posts framed as opinion pieces or thought leadership rarely get cited; add a concrete takeaway in paragraph one

Check 2: Crawl Your Schema Markup

Open Google’s Rich Results Test or run an Ahrefs Site Audit and look at how many of your blog posts carry structured data. In our experience, most blogs have Article schema on some posts and nothing on others, depending on which plugin was active when the post was published. FAQPage schema is almost always missing unless someone added it deliberately.

FAQPage schema is the single highest-leverage schema type for AI citation. When a post contains a properly marked-up FAQ block, AI systems can pull individual question-and-answer pairs directly into their responses. We have seen posts go from zero AI citations to appearing in three or four ChatGPT answer threads within four weeks of adding FAQPage schema to existing content.

In client work we have found that adding FAQPage schema to the ten highest-traffic posts on a blog produces measurable AI citation results faster than any other single intervention — usually within three to six weeks of Google re-crawling.

• Run the Rich Results Test on your five highest-traffic posts first
• Add FAQPage schema to any post with a Q&A section or a list of common questions
• Check that Article schema has dateModified populated — AI systems apply freshness filters
• Speakable schema signals to voice assistants which sections are most answerable; add it to posts targeting conversational queries

Our SEO service includes a full schema audit and implementation pass as part of onboarding — we bring this level of structural attention to every site before writing a single new word.

Check 3: Audit Internal Linking Gaps

Internal linking does two things simultaneously: it passes PageRank between posts, and it tells AI crawlers how your content is organised thematically. Most blogs have accidental internal linking — some posts link to each other because an editor remembered to do it, others are completely isolated islands with nothing pointing to them.

Export your sitemap, run it through Screaming Frog (free up to 500 URLs), and look at the “Inlinks” column. Any post with fewer than two internal links pointing to it is an island. Any post that is a pillar page — a comprehensive guide that should anchor a cluster — should have five or more internal links pointing to it from supporting posts.

• Build a simple link map: list every post and how many internal links point to it
• Identify your three most important pillar posts and make sure every cluster post links to the pillar
• Add contextual links — not “click here” links at the bottom, but linked phrases inside relevant sentences in the body
• Check that your most recent posts link back to relevant older content; Google values link freshness

A Quick Fix for Legacy Content

If you have 50+ posts and manual linking feels impossible, start with the ten most-visited posts. Add two or three contextual internal links to each one, pointing to your strongest pillar content. Then add reciprocal links from those pillars back down to the supporting posts. This targeted intervention usually produces measurable crawl improvements within two or three crawl cycles — and it costs nothing beyond time.

Check 4: Review Title Tags for Dual Optimisation

Traditional SEO title optimisation focuses on the target keyword near the front of the title. AI-SEO adds a second requirement: the title should also read as a question or a clear promise. “AI Content Production for Marketing Agencies” is a reasonable traditional SEO title. “How AI Changes Content Production for Marketing Agencies in 2026” is better for both — it contains the keyword and signals to AI systems that the post will answer a specific how-question.

Go through your title tags and flag any that are noun phrases rather than verb phrases. A noun phrase names a topic. A verb phrase promises an answer. AI systems consistently prefer posts that promise answers over posts that label topics.

• Titles starting with “How”, “Why”, “What”, or “When” perform better in AI citation tests across all major AI platforms
• Include the year in titles for time-sensitive topics — AI systems apply freshness filters aggressively
• Keep titles under 60 characters so they are not truncated in either SERP or AI summary displays

Check 5: Identify Thin and Duplicate Content

Thin content — posts under 600 words that do not fully answer their stated question — is penalised more harshly in AI systems than in classic Google. Google can still rank a thin post if it has strong backlinks; AI systems simply will not cite it. Duplicate content is even more damaging: if you have three posts on the same topic with slightly different angles, AI systems often cite none of them because the topical signal is split.

Use Screaming Frog or Sitebulb to identify posts under 800 words. For each one, ask honestly: does this post fully answer its core question? If not, either expand it to 1,200 or more words with genuine additional substance, or 301-redirect it to the strongest post on that topic and consolidate the content there.

• Merge near-duplicate posts into one authoritative piece rather than trying to rank multiple thin variations
• After merging, 301-redirect the old URL to the new consolidated post permanently
• Update the publication date on expanded posts — freshness matters for both Google and AI citation engines

Our AI content creation service includes a content consolidation step before we write anything new — publishing fresh thin content on top of an existing thin archive compounds the problem rather than fixing it.

Check 6: Verify Excerpt and Meta Description Quality

The meta description does not directly affect AI citation, but it affects click-through rate from traditional search, and engagement signals feed back into rankings over time. More directly relevant: the post excerpt — the first 150 to 200 words — is what AI systems use when evaluating whether a post answers a given query. If your excerpt is vague or scene-setting, AI systems will skip the post even if the body content is excellent.

Review the first 200 words of every post in your top 30 by traffic. Does the opening contain the target keyword, the core answer in plain language, and a reason for the reader to keep reading? If any of those three are missing, the opening paragraph needs a rewrite before anything else will have maximum impact.

• Write meta descriptions that contain the keyword and a specific promise, not generic descriptions of what the post is “about”
• Keep meta descriptions between 140 and 155 characters — truncation in AI-generated summaries follows similar limits
• Avoid opening posts with a rhetorical question that you do not immediately answer — AI systems want the answer first, not a warm-up

Check 7: Audit Heading Structure for Answer-First Formatting

AI systems parse heading structure to understand what sections of a post answer which questions. If your H2 headings are topic labels — “Introduction”, “Background”, “Our Approach” — rather than statement or question-answer pairs, AI systems cannot extract usable answers from them. Rewrite headings as statements or questions that contain the answer inline.

Instead of “Social Media Strategy” as an H2, use “Why Most Social Media Strategies Fail in the First 90 Days” or “The Social Media Cadence That Actually Generates Leads.” The heading itself should tell AI systems what the section answers — not just what it is about.

• Avoid single-word or vague H2 headings — they carry no semantic value for AI parsing
• Use H3 subheadings inside longer sections to break down sub-questions; each H3 creates an additional citation surface
• Do not skip heading levels (H1 to H3 without H2) — this confuses both readers and structured-data parsers
• Check that each post has exactly one H1, which should match or closely mirror the title tag

The Heading Skimming Test

Take your three most-visited posts and try reading only the headings, skipping all body text. Do the headings tell a coherent story? Do they collectively answer the post’s core question? If not, the heading structure needs a rewrite before other changes will compound properly. This test takes three minutes and is one of the most reliable diagnostics we run.

Check 8: Review Image Alt Text and File Names

Image optimisation is often treated as an afterthought, but alt text and file names are read by AI crawlers and contribute to topical relevance signals. An image file named “screenshot-2024-03-12.png” with alt text reading “image” tells crawlers nothing. The same image with a descriptive file name and contextually relevant alt text adds a small but real topical signal to the post.

Run a bulk alt text audit with Screaming Frog (Images tab, filter for missing alt text). Fix every missing alt text on your top 20 posts first. For existing alt text, check that it describes what the image shows and includes a relevant keyword naturally — not stuffed in artificially.

• Rename image files to descriptive slugs before uploading — you cannot rename them in WordPress after upload without breaking URLs
• Keep alt text under 125 characters — screen readers truncate at this point and some crawlers follow the same limit
• Avoid duplicating identical alt text across multiple images within the same post

Check 9: Check Page Speed and Core Web Vitals

Page speed has been a Google ranking factor for years. Its impact on AI citation is indirect but real: slow pages get crawled less frequently, which means freshness signals degrade over time. More directly, a user who bounces from a slow-loading post sends a negative engagement signal that erodes rankings gradually.

Run your top 20 posts through Google PageSpeed Insights. Focus on Largest Contentful Paint (LCP) under 2.5 seconds and Cumulative Layout Shift (CLS) under 0.1. These two metrics account for the majority of user experience complaints on content sites and are the ones Google’s quality raters pay closest attention to.

• Convert all images to WebP format — this is the single highest-impact speed fix on most content blogs
• Defer non-critical JavaScript — most WordPress themes load scripts that are not needed on first render
• Enable browser caching for static assets — any reputable WordPress caching plugin handles this automatically without custom code

Check 10: Audit External Links for Authority and Freshness

External links matter in both directions. Links pointing to high-authority, current sources signal to AI systems that your post is grounded in real, verifiable data. Links pointing to dead pages or outdated sources — a 2018 study cited as current evidence — undermine credibility. AI systems increasingly apply freshness filters to source material, and a study on user behaviour from five years ago carries very little weight in a 2026 answer.

Export all external links from your posts using Screaming Frog. Check for 404 errors on outbound links. Flag any links to sources older than three years for time-sensitive topics and replace or remove them. This is tedious but it matters — broken outbound links are a surprisingly common audit finding on blogs with more than two years of history.

• Link to primary sources where possible — academic papers, official platform documentation, government statistics
• Avoid linking to competitor blog posts as sources; link to the original study they cited instead
• Use rel=”nofollow” on affiliate links and sponsored content — this is a crawl hygiene signal, not just a disclosure requirement

Check 11: Verify Canonical Tags and Index Status

Canonical tag errors cause duplicate content issues that both Google and AI crawlers penalise. A post canonicalised to itself is correct. A post canonicalised to a different URL is telling crawlers to ignore it. A post with no canonical tag leaves ambiguity that different crawlers resolve differently. Run your blog through Screaming Frog and filter the Canonicals tab for any anomalies — they are usually easy to fix once you can see them.

Also check your robots.txt and any noindex tags. We have audited client sites where entire tag archives, author archives, and search result pages were indexed, diluting crawl budget across hundreds of thin auto-generated pages. Every crawl request spent on a tag archive page is a request not spent on a substantive post that could be cited by AI systems.

• Block tag and author archive pages from indexing via robots.txt or noindex meta tags in your SEO plugin
• Ensure pagination pages are canonicalised to the first page or handled properly with pagination signals
• Check that your sitemap.xml only includes URLs you want indexed — many WordPress plugins include everything by default, including drafts and private pages

Check 12: Score Each Post for AI Citability

The final check is a holistic citability score for each post. We use a simple five-point rubric: one point each for (1) a direct answer appearing in paragraph one, (2) FAQPage or Article schema present and valid, (3) at least two internal links pointing to the post, (4) headings framed as questions or direct statements rather than topic labels, and (5) the post updated within the last 12 months. Any post scoring two or below goes on a priority rewrite list. Posts at three or four get targeted fixes. Posts at five are your citation-ready assets.

Apply this score to your top 30 posts by traffic. You will usually find that your highest-traffic posts cluster in the two-to-three range — they rank because of backlinks and historical domain authority, not because they are AI-citation-ready. Those are your highest-leverage targets, because improving their AI citability adds an entirely new traffic channel on top of existing search traffic rather than replacing it.

• Build a simple spreadsheet with one row per post and a column for each of the five scoring criteria
• Prioritise posts where the target keyword has high search volume and clear question intent — these are most likely to surface in AI-generated answers
• Schedule rewrite sprints rather than trying to fix everything in one week — four posts per week is sustainable and the results compound over time
• Re-run the full audit every six months; AI citation patterns shift as platforms update their training data and ranking logic

Where to Start If This Feels Overwhelming

Twelve checks on a blog with 50 posts can feel like a month of work. In practice, 80% of the impact usually comes from three interventions done in order: adding FAQPage schema to your top-10 posts, rewriting the opening paragraph of any post that does not answer its core question directly in the first 150 words, and fixing internal links so your pillar posts are properly connected to their cluster posts. Do those three things first and you will see measurable results before you finish the full audit.

If you want to go deeper on the structural content patterns that AI systems respond to, our post on how to get cited by ChatGPT covers answer-first writing, TL;DR blocks, and the formatting approaches we use on our own posts and in client work.

If you have completed the audit and want a team to execute the fixes systematically, we work with a small number of clients on exactly this kind of structured AI-SEO content improvement — methodical, without the churn of publishing more thin content on top of an already-thin archive. Visit our contact page to start a conversation. We will tell you honestly whether we are a fit before either of us commits to anything.