From 55 to 82 in 6 Weeks: HelpSquad's AEO Rank Iteration Story

HelpSquad, a live-chat and customer-support software brand, started a focused AEO Rank engagement in early April 2026. The instant-engine baseline on 2026-04-09 was 55. Six weeks later, on 2026-05-24, the same engine returned 82. That 27-point gain landed HelpSquad in the top 32 of our 13,385-site corpus — a cohort containing only 0.24 percent of all scored sites. The path between those two scores was not a clean upward line, and the regression points in the middle are part of the story.

The starting score: 55 and a flat-but-decent pillar profile

A 55 on AEO Rank is above the corpus median of 47 but below the p75 of 53 — the engine was reporting a site already doing better than half the corpus, but well outside the citation-ready tier above 70. The HelpSquad before-audit pillar profile was unusually flat:

Pillar	Score (v1, 2026-04-09)	Corpus average
AI Discovery	7.4	4.1
Trust & Authority	7.4	3.7
Content Structure	6.8	4.8
Technical Foundation	6.2	5.3
Answer Readiness	7.9	5.7

Every pillar was at or above the corpus average — there was no obvious weakest link to fix. The audit’s specific gaps were criterion-level inside the otherwise decent pillars: missing Speakable Schema, weak comparison-table coverage on product pages, and an FAQ block that only existed on a handful of high-traffic pages rather than across the catalog.

The iteration curve: four versions, two regressions, one sustained peak

The full instant-engine progression from aeo_audit_versions:

Version	Date	Score	Notable change
v1	2026-04-09	55	Baseline
v2	2026-04-13	78	+23 in four days from first AEO Rank fixes
v3	2026-05-04	60	Regression: schema validation issue on new pages
v4	2026-05-09	55	Further regression: dropped to baseline
v5	2026-05-22	84	Peak after Content Structure rebuild
v6	2026-05-24	82	Sustained at 82

The v3 and v4 regressions are the most instructive part of this case study. The team had introduced new product comparison pages between v2 and v3 with malformed schema markup — broken Product JSON-LD on a handful of templates dragged Trust & Authority and the comparison-table criterion down. Re-running the audit caught the regression at v3. Fixing the schema and validating with the Rich Results Test returned the score to v5’s peak.

The pillar diff that explains the 27 points

Looking at the v1 to v6 pillar deltas tells a clean story:

Pillar	v1 (55)	v6 (82)	Delta
AI Discovery	7.4	7.9	+0.5
Trust & Authority	7.4	7.0	-0.4
Content Structure	6.8	9.1	+2.3
Technical Foundation	6.2	7.3	+1.1
Answer Readiness	7.9	8.4	+0.5

Content Structure carried the lift. The 2.3-point pillar move maps to two specific criterion changes: Table and List Extractability moved from a near-zero baseline to a strong score once HelpSquad shipped HTML comparison tables across every product line (one table per page, three to five columns, thead and tbody properly marked), and Q&A Content Format moved into the top tier once FAQ blocks with FAQPage schema were attached to every product page rather than just three or four.

Trust & Authority slightly regressed (-0.4). That is real and worth flagging: a content sprint that lifts other pillars can inadvertently dilute T&A if new pages ship without named-author attribution or visible dates. HelpSquad’s v6 is a strong overall score even with this minor pillar regression, but the team’s next iteration is targeting T&A specifically.

What “iterative” actually looks like

The cleanest takeaway from the HelpSquad story is that AEO Rank work is iterative, not linear. The v2 spike to 78 was real but partly luck — early easy wins. The v3 and v4 regressions were real too, and they corrected misplaced confidence that “we are done at 78.” The v5 peak at 84 came after the team systematically caught and fixed every schema validation issue introduced between v2 and v3. The v6 settle at 82 is the steady state.

For teams approaching AEO Rank work, the lesson is to re-audit frequently. The engine catches structural regressions you would not notice from page reviews alone. HelpSquad’s v3 score told the team something was wrong before any customer-facing impact, and the fix was a fifteen-minute schema correction.

How We Tested

All six audit versions are publicly visible at https://audit.aeocontent.ai/helpsquad-com and the underlying scorecards are stored as immutable rows in the aeo_audit_versions table on our production database. Every score in this article was queried from that table on 2026-05-29 — no estimates, no proxies, no engagement-internal data that is not also in the public corpus.

Pillar averages are computed by taking the weighted average of every criterion’s score within each pillar, where the weight is the criterion’s contribution to the pillar’s portion of the overall AEO Rank. The corpus comparison column (“corpus average”) uses the same weighted-average method against every scored domain in aeo_published_audits.

The engine and the scoring schema (instant v5.0, 48 criteria) were held constant across all six versions, so the deltas reflect site changes rather than engine changes. Cross-engine deltas (for example, HelpSquad’s Claude-engine score of 77 from February 2026) are out of scope for this article — they are honest but not directly comparable to instant scores because each engine evaluates a different signal set.

What to take from this

If your AEO Rank is in the 50-to-60 band today, the HelpSquad path is replicable. Identify one pillar where your score lags the corpus average. Ship the structural changes that target the underlying criteria. Re-audit weekly so regressions surface before they harden. Read the AEO Rank methodology for the criterion-to-pillar mapping, or run a site audit to see your own pillar profile and the criterion-level gaps the engine flags.