Google Drops num=100: Reset AEO Measurement Around AI Answer Visibility

What changed and why it matters\n\nOn September 18, 2025, Google confirmed the results per page parameter is not supported, ending the long standing num=100 workaround that many tools used to collect deep results. This coincided with visible drops in Google Search Console impressions and disrupted third party rank tracking. See the confirmation in this report: Google ends results per page parameter.\n\nIndependent analyses show widespread impact on reported visibility. One dataset covering 319 properties found 87.7% saw impression declines and 77.6% lost unique ranking terms after the change. Review the analysis: num=100 impact data across 319 properties.\n\nIn parallel, Google is pushing AI answer experiences like AI Overviews and Chrome’s AI Mode, where answers appear with citations and links. To capitalize on those surfaces, learn how to optimize for Chrome Gemini and build a discovery strategy using our Session Engine Optimization guide.\n\n## The measurement reset: from rank to answers\n\n- Retire average position as a primary KPI. It depends on paginated SERPs and deep result collection that are now unreliable and expensive.\n- Replace with answer centric metrics that reflect how users actually see and click brands inside AI answers across engines.\n\n## Two core KPIs to adopt\n\n### 1) Answer Impression Share (AIS)\n\n- Definition: The percentage of answer impressions for your tracked queries in which your brand is cited or linked inside the AI answer unit.\n- Formula: AIS = cited answer impressions for brand ÷ total answer impressions for tracked queries.\n- Use cases: Benchmark brand presence across Google AI Overviews, Chrome AI Mode answers, Bing Copilot, Perplexity, Brave, and others. Segment by engine, intent cluster, market, and device.\n\n### 2) Citation Depth\n\n- Definition: How prominently the brand appears inside the answer unit.\n- Track: first citation index, number of citations, and whether the citation renders above the fold in the answer card.\n- Metrics: first citation position, total citations per answer, an above fold flag, and a weighted Citation Depth Score where earlier and above fold placements score higher.\n\n## Lightweight measurement blueprint\n\n### A) Browser sampling for answer visibility\n\n- Queries: Maintain a representative basket per market and intent cluster. Include brand, category, and problem led queries. Refresh quarterly.\n- Sampler: Use real browser automation with human like settings such as headless Chrome with a real user agent, signed out, U.S. IPs, and locale pinned. Respect robots.txt and rate limits.\n- Rendering: For Google, measure classic web results and detect AI answer panels. Parse citation blocks, extract domains, URLs, order, and visibility flags. For Bing Copilot, Perplexity, and others, detect answer sections and outgoing citations similarly.\n- Output fields: timestamp, engine, query, country, device, answer_present, citations [domain, url, order, above_fold], answer_type, and follow up affordances.\n- Cadence: Daily for head terms, weekly for long tail. Target 95% confidence in AIS within a ±3 to 5 point margin by adjusting sample size.\n\n### B) First party logs to validate real exposure\n\n- Tag cited URLs with a persistent UTM schema and a lightweight landing script to record referrer, navigation type, and viewport at arrival. Do not block preload or prerender traffic, but label it.\n- Build heuristics to attribute answer sourced visits such as referrers from google.com with answer pane link patterns, Bing Copilot answer link refs, and explicit redirectors used by answer engines. Validate with controlled tests and annotation.\n- Compare session rates from answer engines with sampled AIS to catch undercounting or over attribution.\n\n### C) Scoring and reporting\n\n- Answer Visibility Index (AVI): Combine AIS with normalized Citation Depth for a single number per engine and cluster. Example: AVI = AIS × normalized depth.\n- Coverage: Percent of tracked queries that trigger any answer unit, and percent where your brand is present.\n- Freshness: Median age of your cited content in answers. Refresh key pages if stale citations dominate.\n- Competitor share: AIS by top competitor domains to spot displacement risk.\n\nFor cost and governance strategies that align crawling and licensing with AI answers, explore the RSL robots.txt licensing model.\n\n## Governance and quality controls\n\n- Separate measurement sampling from production crawling. Enforce strict rate limits and honor engine terms.\n- Use reproducible query lists with versioning, and store raw HTML snapshots to reparse as answer UIs evolve.\n- Monitor drift. If engines change UI or markup, rerun validation suites before publishing metrics.\n\n## Transition plan for teams\n\n- Week 1 to 2: Freeze average position as a legacy diagnostic. Stand up the sampler and baseline AIS and Citation Depth for your top 200 queries across Google and one secondary engine.\n- Week 3 to 6: Expand coverage to your full strategic set. Wire AIS and AVI to your KPI dashboards. Add competitor share.\n- Week 7 to 12: Tie AIS movements to content programs. Ship briefs to improve documents that are frequently cited but low depth, and create net new pages for unanswered intents.\n\n## What good looks like\n\n- AIS above 40% in branded and near branded clusters and rising share in key unbranded problem led queries.\n- Citation Depth improving with the first citation appearing in the first one to two links of the answer card and above the fold.\n- Consistent correlation between AIS gains and answer sourced sessions in analytics, confirming you are measuring exposure that users actually see.\n\n## Bottom line\n\nThe num=100 era is over and deep SERP rank obsession should give way to answer visibility. Treat AI answer surfaces as new distribution and measure citation share and prominence across engines. That is the durable way to understand and grow search exposure after September 2025.