What moved, and what didn't.
A live re-run of the measurable findings from the May 2026 AEO/SEO audit, twenty days after delivery. The page-level polish started. The fix that gates everything did not.
In the twenty days since delivery, two of the cheapest fixes shipped — the AEO product page now reads coherently (title and H1 finally agree), and the /robots.txt redirect loop is gone. The single most consequential finding did not move: Bingbot, MSNBot and Yandex are still returning HTTP 403 at the WAF, including on /robots.txt itself.
Because ChatGPT and Microsoft Copilot both read Bing's index, the retrieval door is still shut. The 4% own-domain citation ceiling is mechanically locked until that one rule comes off. Net visibility position: effectively unchanged.
| Finding | Baseline (v1) | Now (28 May) | Status |
|---|---|---|---|
| Bing crawlers blocked at WAF§04 · Finding 01 | Bingbot / MSNBot / Yandex → 403 on homepage and /robots.txt. AI bots → 200. |
Bingbot / MSNBot / Yandex → 403 on homepage and /robots.txt. AI bots → 200. Googlebot → 200. |
No change |
| robots.txt redirect loop§04 sub-issue | Browser UA: /robots.txt → /robots.txt/ → /Util/Errors/Error405/, self-looping. |
Browser UA: /robots.txt serves 200 OK directly, zero redirects. |
Fixed |
| AEO page title/H1 split§06.e | title "AEO Visibility — Siteimprove" vs generic value-prop H1. |
title & H1 both "Advanced AEO Insights to Elevate Your AI Visibility" — matched. | Fixed |
| Title/H1 — other pages§06.e | Home, /platform/, /accessibility/ each split title vs H1. | Home ("Agentic Content Intelligence" vs "Where accessibility meets performance"), /platform/, /accessibility/ still split. | Partial |
| Schema coverage§06.a | Home: Organization + WebSite only. AEO page: template-only. Comparison pages: template-only. | Identical. Home still no SoftwareApplication / Product / FAQPage. AEO page still 0 SoftwareApplication, 0 FAQPage. | No change |
| Duplicate product-card label§06.c | "Accessibility agents help you review before you publish" ×4 under all cards. | Same label still rendered ×4. | No change |
| Comparison hub§06.d | 7 pages. Deque, Conductor, TPGi missing. | Still 7 pages (Level Access, Silktide, Acquia, BrightEdge, SEMrush, Matomo, GA4). Deque, Conductor, TPGi still missing. | No change |
| llms.txt§06.b | Hard 404. |
No file. 301 → /Util/Errors/Error404/ (soft-404, served 200). Substance unchanged; hygiene slightly worse. |
Watch |
| AI citation & brand memory§05 · Finding 02 · directional | 4% own-domain share. 0/37 brand-memory branches mention AEO/agentic. | Third-party sources (TestParty, G2, Webability, A11y Pulse) still own "Siteimprove alternatives"; own pages absent up top. Brand memory unchanged — AEO launch (Apr 20) too recent for training data. | No change |
What shipped
Both fixes that landed are page-level and low-effort, and both were flagged as easy wins in v1. The AEO product page at /platform/seo/aeo-visibility/ now carries one coherent signal end to end — the <title> and <h1> say the same thing, which is exactly the §06.e remediation. The /robots.txt redirect loop that previously bounced through an internal error endpoint is also resolved under a browser user-agent. These are real improvements. They are also the two cheapest items on the list.
What didn't, and why it's the whole story
The headline finding has not moved. Bingbot, MSNBot and Yandex still return 403 at the WAF — on the homepage and on /robots.txt itself, which per RFC 9309 §2.3.1.3 a crawler reads as "do not crawl." ChatGPT and Copilot are both Bing-fed; with the index they read going stale, they keep falling back to training-era memory. Every downstream lever in the audit — schema, comparison-hub depth, fresh content — only pays out once crawlers can reach the pages. Until this one rule comes off, the rest is effort spent behind a closed door.
The schema gap, the duplicate card label, and the three missing comparison competitors (Deque, Conductor, TPGi) are all unchanged. None is the highest-leverage item, but the duplicate label is a five-minute edit that hasn't happened, which suggests the on-site backlog hasn't been picked up beyond the single AEO page.
The citation layer (Finding 02)
Re-run directionally — a one-engine memory test plus the live grounded SERP, not the full five-engine sweep that produced the 4% figure. Both halves read the same as the baseline. The grounded comparison layer is still owned by the same third parties: TestParty's "Siteimprove alternatives" page, G2's competitor list, Webability.io and A11y Pulse all still rank for the buying queries, and siteimprove.com's own comparison hub still does not surface among them. Pricing is still set by outsiders — "$20K–$30K/year" now circulates in place of TestParty's earlier "$15K–$50K+," and the first-party pricing page the audit recommended still does not exist. Brand memory is unchanged, which is expected: the "Advanced AEO Insights" launch shipped on 20 April 2026, far too recently to have entered any model's training data, and twenty days is not a window in which ungrounded memory moves.
The one genuinely new signal is on the grounded side: the April launch and the agentic-platform expansion drew real press coverage (PR Newswire, Morningstar, MarTech Series, CMSWire, KMWorld). That is fresh third-party authority that can re-anchor the entity over time. It is also undercut by the same block as everything else — the two engines that read Bing still cannot reach siteimprove.com to corroborate it from the source.
Twenty days in, the team fixed the two cheapest page-level items and left the WAF rule that gates ChatGPT and Copilot retrieval untouched — so own-domain AI citation share has no mechanical reason to have moved off 4%.
The technical findings (01 and 03) were re-run twice against the live origin from a single IP on 28 May 2026, with identical results on both passes. The tests covered a representative crawler set — Bing, MSN and Yandex variants against the AI bots, OAI-SearchBot, Google-Extended and Googlebot — on the homepage and /robots.txt; they did not reproduce the full 147-bot sweep, so the v1 "85% health score" is not restated here. On-site checks read the rendered HTML and JSON-LD of the homepage, /platform/, /accessibility/, the AEO product page, the comparison hub and the Silktide comparison page. Finding 02 was re-run directionally only — a single-engine memory check plus the live grounded SERP for the buying queries — not the full five-engine, 30-prompt sweep or a Brand Radar pull that would reproduce the exact 4% number; the v1 sweep was itself flagged as directional rather than statistical, and this pass is held to the same standard.
// crawler re-test — www.siteimprove.com, 28 May 2026 Bingbot home=403 robots.txt=403 MSNBot home=403 robots.txt=403 YandexBot home=403 robots.txt=403 Googlebot home=200 GPTBot home=200 robots.txt=200 OAI-SearchBot home=200 ClaudeBot home=200 robots.txt=200 PerplexityBot home=200 robots.txt=200 Google-Extended home=200 // resource checks /robots.txt (browser UA) 200 OK, 0 redirects ← loop fixed /llms.txt 301 → /Util/Errors/Error404/ ← still no file /sitemap.xml 301 → /sitemap/ (text/xml, 139 KB) // title vs H1 /platform/seo/aeo-visibility/ title == H1 ← fixed / title != H1 /platform/ title != H1 /accessibility/ title != H1 // schema @types still present on audited pages homepage Organization, WebSite (no SoftwareApplication/Product/FAQPage) aeo-visibility WebSite/SearchAction only (no SoftwareApplication/FAQPage) comparison hub WebSite/SearchAction only
Caveat carried from v1: tests run from a single IP. WAFs sometimes block by region or ASN, so the Bing 403 should be confirmed from a second region before stating the block as universal — though the pattern (every Bing/MSN/Yandex variant blocked, no AI bot blocked) makes a per-IP false positive unlikely.