The AEO Audit I Ran on My Own Webflow Practice This Week

I sell AEO audits to client prospects. I have been doing it for almost a year. This week I ran the same audit on my own pravinkumar.co. The findings were uncomfortable. Some were predictable. A few were genuinely surprising. The temptation was to fix everything immediately and not write about it. I am writing about it because the public version of an honest audit is more useful than a private one, both for me and for any other Webflow Partner reading along. This piece walks through what I found, what I changed, what I left alone for now, and what the exercise taught me about the gap between selling AEO and living it.

Why Did I Decide to Audit My Own Site This Week?

The trigger was a client conversation last Friday. A founder I had been talking to about a content retainer asked, casually, whether I had run the same audit on my own site. I gave a confident answer that turned out to be roughly half-true. I had run pieces of the audit. I had not run the full version end to end. The honest version of the answer would have been "not in the last six months" and the founder probably knew that from the way I phrased it. The exchange ended politely. The proposal stayed open. I went back to my desk and ran the audit Sunday morning.

The deeper reason is that publishing 217 articles in 25 days is the kind of activity that produces compounding metadata drift. The tags, the meta descriptions, the canonical URLs, the structured data, the internal linking patterns. Each individual article gets quality attention at write time. The cumulative coherence across the corpus is something nobody is checking unless somebody explicitly checks. That somebody is me. I had not been doing the check. I covered the publishing discipline in my six months daily publishing piece.

What Are the Audit Steps I Actually Run for Clients?

The audit has six sections. Crawl accessibility for AI bots, including robots.txt configuration and Cloudflare-layer bot blocking. Schema markup completeness across article, breadcrumb, and FAQ types. Citation-readiness of individual articles, including the answer block density and the named entity count. Internal linking topology, including the orphan-page check and the inbound link distribution. Content freshness, with a flag on any article older than ten months. Cross-engine visibility, with manual checks of how five sample articles render in ChatGPT Search, Perplexity, Claude Search, and Google AI Overviews.

Each section produces a finding sheet with red, yellow, or green status, plus a specific remediation step for any non-green finding. The audit takes roughly four hours when run on a 50-page client site. Mine has 217 pages, which extended the audit to about seven hours spread across two evenings. The volume itself was a finding. A practice that has scaled past 200 articles needs continuous AEO monitoring, not periodic audits. I had not built that monitoring. I covered the related discipline in my tracking AI search visibility piece.

What Did the Crawl Accessibility Section Find?

This was the most surprising section. My robots.txt allows the major AI search crawlers correctly. Cloudflare's bot management was not silently blocking AI crawlers above my robots.txt, which I checked manually. The surprising finding was a pattern of Claude-User and Claude-SearchBot fetches concentrated on five specific articles in the AI category, with much lower fetch density on the design and personal categories. The pattern suggests Claude weighs my AI category content as the most authoritative on the site, which is good news, but the design and personal categories are underweighted, which means article-level effort is not translating to category-level visibility.

The remediation is to strengthen internal linking from the AI category into design and personal articles, which routes Claude's existing crawl interest into the underweighted sections. The change took about an hour for the highest-traffic AI articles and produces compounding visibility across the whole corpus over the next two months. I had been writing each article's internal links forward but not consistently routing crawl interest backward into older content. The audit caught the asymmetry that month-by-month attention misses.

What Did the Schema Markup Section Find?

Article schema and breadcrumb schema are both in place across the corpus. The finding was that FAQPage schema is missing on most articles even though most articles contain question-based H2 headings that would qualify. Adding FAQPage schema to existing articles is a Webflow Custom Code job that scopes to each blog post template, which I had been deferring because the content rules I follow generate H2s as questions but the body answers do not always match the cleanest FAQPage format.

The remediation here is to write a small Cloudflare Worker that injects FAQPage schema into the response for blog post pages, generating the JSON-LD from the page's H2 and following-paragraph structure. The Worker is a four-hour build that improves AEO citation readiness across all 217 articles plus every future article. I had been adding the schema piece manually on a few articles, which is the worst of all worlds. Either build the system once or skip the schema entirely. Doing it inconsistently produces noise without compound benefit. I covered the related schema discipline in my schema markup types piece.

What Did the Citation-Readiness Section Find?

Answer-block density is strong on articles written in the past 60 days. The articles older than 90 days were written before I had fully internalized the 40 to 60 word answer-block discipline at the top of each H2, and the older articles trail in extraction-friendliness as a result. The remediation is to refresh the older articles with proper answer blocks. The volume is roughly 80 articles, which is too many to refresh manually in a single session.

The plan I landed on is to refresh five articles per week as an ongoing maintenance cadence, prioritized by current Google traffic and citation appearances. Five per week clears 80 articles in 16 weeks, which is the right pace for a one-person practice. The deeper finding is that AEO discipline, like any quality bar, takes time to fully internalize, and the articles written during the learning curve will need a second pass once the discipline is set. I had been treating each article as a one-shot deliverable rather than as part of a corpus that needs periodic recalibration. I covered the related freshness discipline in my AI citation decay piece.

What Did the Internal Linking Section Find?

The orphan-page check surfaced 14 articles with zero or one inbound link from other articles in the corpus. Some of these are intentional, like very specific niche pieces that do not warrant being linked to from generalist articles. Most are unintentional, where the article exists but the older articles it should connect to do not yet point at it because they were published before the new article existed. The bidirectional backlink debt I had been tracking turned out to be larger than I thought.

The remediation is the bidirectional backlink sweep that has been on the open editorial debt list across three weeks of publishing. The sweep adds a backlink to each new article from two or three older articles in the corpus that thematically connect. The volume is roughly 160 link additions across 90 articles. The work is unglamorous, takes three to four hours per session, and would mostly be done by an AI agent if I were running this for a client. Running it for myself, I caught up on it Sunday afternoon and am committed to including it in every future publish day rather than letting it accumulate again.

What Did the Content Freshness Section Find?

Nothing on the site is older than ten months because I started publishing in earnest at the start of 2026. The freshness flag is therefore green across the whole corpus today. The forward-looking finding is that the earliest articles cross the ten-month threshold around November 2026, at which point the freshness flag turns yellow. The remediation is to plan a scheduled refresh sprint for those early articles in October 2026, which gives a one-month buffer before AI systems start de-weighting them.

The deeper insight is that freshness is a calendar problem more than a content problem. The articles do not get worse over time. The signals AI systems use to decide which content to cite shift as content ages. A 12-month-old article on a topic that has not changed substantively is still factually right, but it loses citation share to newer articles on the same topic by other authors. The defense is to refresh the article with a published date update and a few lines of new context that ground it in the current quarter. I covered the related discipline in my content refresh strategy piece.

What Did the Cross-Engine Visibility Spot-Check Find?

I ran five sample queries across ChatGPT Search, Perplexity, Claude Search, and Google AI Overviews. The queries ranged from broad ("webflow ai partner") to specific ("webflow logic forms vs zapier 2026"). The findings were uneven. Perplexity and Claude Search cite my articles with reasonable consistency on the specific queries. ChatGPT Search cites less often and tends to prefer larger publication sources. Google AI Overviews surface my articles intermittently with no clear pattern.

The remediation is harder for this section because the work is less under my direct control. Engine-level citation patterns evolve based on each model's training data and retrieval logic, which the Webflow Partner cannot change. What I can change is the content shape that each engine seems to prefer, which means the comparison-format pieces I have been publishing are likely to perform better than the personal-essay pieces in the AI search surfaces, even though both serve the practice. The strategic implication is to keep doing both, but not to confuse personal-essay performance with AI search performance. They serve different goals and the metrics for each should stay separate.

What Did I Change Immediately, and What Did I Defer?

Three changes shipped this week. The bidirectional backlink sweep across the 160 missing links. The Cloudflare Worker for FAQPage schema injection on blog post pages. A new internal linking rule that every published article must have at least three inbound links from existing articles within seven days of publish, enforced by a small script that flags violations.

Four changes are deferred. The 80-article answer-block refresh, scheduled at five per week starting next Monday. The freshness refresh sprint, scheduled for October 2026. A monitoring dashboard that watches AI search citation patterns weekly, planned for build in the next two weeks. A schema-completeness check for new articles at create time, planned to be added to the publishing script alongside the existing slug-collision check. The deferred items are not less important. They are sequenced so the immediate fixes ship without crowding out the maintenance cadence the practice needs to sustain over months.

What Is the One Honest Lesson From Running This Audit on Myself?

The lesson is that the gap between selling AEO and living it is real, even when the seller is genuinely committed to the discipline. The volume of work that goes into running a publishing operation at any meaningful pace produces drift that quality intent alone does not catch. The practice that sells AEO needs to operate AEO with the same rigor it sells. Otherwise the sale is rhetorical rather than evidenced, and the prospects who ask the casual question on Friday afternoon will continue to politely close their proposals.

For solo Webflow Partners reading this, the recommendation is to run the audit on your own site this quarter before you sell another client engagement. The audit will surface things you did not know about your own work. The findings will be uncomfortable. The remediation will take several weeks. The compounding effect on credibility is the largest growth investment a one-person practice can make in 2026. I covered the rhythm that supports this kind of self-discipline in my daily habits piece, and the AEO basics in my whether to offer GEO service piece.

If you are running a Webflow practice and want to talk through what an AEO audit on your own site might surface, drop me a line and tell me when you last reviewed the cumulative metadata across your full corpus. Let's chat.