Two of five phases are done. ATB is cited inside an AI Overview zero times.

Full roster detail, including domain_relevance and the per-competitor gap tables, in keyword-brief.html §04 and §09.

Full-coverage motivation tagging (0 ambiguous rows, up from 30% coverage in the prior revision) reversed an earlier, partial-data reading that looked roughly sell/buy balanced. ATB’s presence rate is roughly even between the two sides — the buy-side pool is simply much larger. Detail in keyword-brief.html §06–07.

The single most-repeated case: “cpa firms for sale in california” (volume 70) is cited inside the AI Overview by four different competitors — go2bbi.com, accountingfirmsold.com, accountingpracticesales.com, and poegroupadvisors.com — each at AI Overview position 1, while ATB ranks organically at positions 4–60 across four different pages and is cited by none of them. The two leading citing domains are the same two zero-organic-overlap, client-named domains that lead breadth above — a domain invisible to keyword-overlap discovery can still be the domain an AI Overview cites on ATB’s own core sell-side terms. Source: gaps/gap-ai-overview.csv.

This is a wider pattern than the AI Overview gap alone — the same zero holds for local pack, featured snippets, and People Also Ask. Source: gaps/gap-serp-features.csv (592 rows).

The pattern already named in Phase 1 — most competitors’ keyword gap isn’t cleanly on-category — holds again at the page level. Sources: gaps/gap-keywords.csv, gaps/gap-pages.csv.

Method — the same judgment Phase 1 already established for all-keywords.csv, applied fresh to Phase 2’s larger gap sets (most of which were never run through Phase 1’s classification — a different, larger pull). Six ordered rules, first match wins, same discipline as the motivation rule set:

• 1. Brand-collision. The keyword textually matches a roster competitor’s name/domain root and ranks for ≤2 distinct domains across the whole gap-keywords.csv set — real ranking concentration, not just a text match, decides this (see QA below for why the text-only version of this rule was wrong on the first pass). → off-category.
• 2. Phase 1’s base gate, reapplied unchanged: on-category requires both a transaction term (sale, buy, merger, valuation, succession, broker, roll-up, financing, acquisition, transition, exchange) and an accounting-entity term (practice, firm, CPA, accounting, accountant, bookkeeping) in the same keyword. Anything short of both is off-category here — exactly Phase 1’s own dropped.csv treatment (off_category_no_transaction_entity_match), not a new judgment call.
• 3–5, applied only to keywords that passed rule 2 (both terms present): payroll content → off-category · audit-lead-gen (financial-statement/compliance audit services, a different service line) → off-category · generic tax-prep/compliance content (filing, IRS forms, tax calculators, crypto/foreign-exchange tax topics) → off-category.
• 6. Out-of-state / out-of-country regional-hub content. A non-CA US state name in the keyword itself, or the competitor’s own ranking URL is a non-CA state/Canada regional-hub page or an individually geo-tagged listing page (e.g. .../united-states/northeast/new-york/, .../canada/ontario/, or a listing slug decoding to “…, fl”) — verified from the URL’s own geography, the same standard already used for the accountingpracticesales.com New York hub page above. → off-category.
• Anything clearing all six rules → on-category.

Two genuinely close calls, resolved on evidence rather than forced either way, both caught in QA: “cpa sales practice” and “accounting practice sales [state]” variants text-match a competitor brand (cpasales.com, accountingpracticesales.com) but each ranks across 5–7 unrelated domains in the gap dataset, not just the one domain the brand belongs to — real ranking behavior overrules the text match, so these were kept on-category (or, for the state-tagged variants, correctly routed to rule 6 instead). An earlier pass of rule 1 used text-matching alone and wrongly excluded both as brand-collision; fixed by gating the rule on ranking concentration (≤2 domains) before accepting a text match, the same domain-diversity check now folded into the rule itself. prohorizons.com/pacific/ was kept on-category rather than flagged out-of-state despite the page-level region name, because Phase 1 already documented prohorizons.com as CA/OR/WA-focused and one of its own on-category gap keywords on that exact page is “ca practice sales” — real evidence of California relevance, not a forced call either way.

Off-category rule breakdown, for transparency — gap-keywords.csv: no transaction/entity match 6,505 · out-of-state/out-of-country regional hub 312 · brand-collision 22 · payroll 3 · generic tax-prep 2 · audit-lead-gen 0 (no keyword that cleared the base gate also matched this rule). gap-pages.csv: no transaction/entity match 1,208 · out-of-state/out-of-country 25 · brand-collision 8. New files: gaps/gap-keywords-on-category.csv (762 rows) and gaps/gap-pages-on-category.csv (87 rows), each carrying a new on_category_rule column stating which rule matched. The original gap-keywords.csv / gap-pages.csv are untouched — still the full/general reference, unchanged row-for-row.

Method: for each cluster, its keywords were intersected against the exact 602-row keyword set atbcal.com’s own domain/keywords pull returned (the same set Phase 2’s 2.3/2.4 sge with_link/without_link pulls were run against) to get “ATB ranks organically.” Of that intersection, each keyword’s own serp_features field (cross-validated against the saved sge without_link raw file — both methods agree exactly, 90 of 90 keywords) determined whether it triggers an AI Overview. Since ATB’s global sge with_link pull already returned a genuine, structurally confirmed empty set (0 of 602 rows, logged in RUN-LOG.md, Phase 2 step 2.4), every cluster subset of that same 602-row set is mathematically guaranteed to be 0 as well — this table reports that per cluster rather than leaving it assumed. 8 of 11 clusters have at least one AI-Overview-triggering keyword ATB ranks for; in all 8, the citing count is 0. Practice listings (11) and ATB’s own branded cluster (6) carry the most such keywords — the largest exposure, and still zero. Separately, and unrelated to citation: the two clusters with the highest cluster-wide AI Overview trigger rate (blog/news, succession/merger) are also where ATB’s organic presence is thinnest or entirely absent. Full method and the per-cluster script output logged in RUN-LOG.md; the un-cross-referenced 11-row table (cluster-wide share only) remains in competitive-map-summary.md §5, unchanged.