SEO Reference - Thresholds, Evidence, and the 2026 Search Landscape
Your SEO instincts are built on CTR benchmarks and ranking signals that no longer match what Google actually does. The numbers changed faster than most guides updated.
The 2024 Google API leak confirmed signals Google had publicly denied using for years: click behavior via NavBoost, domain authority, Chrome browsing data. That changed how Wire's thresholds were set. But the leak is only one layer. AI Overviews have since cut position-1 CTR by more than half, and Google now rewrites 76% of title tags. If you're reading audit output and the numbers don't match your expectations, the benchmarks you're comparing against may be the problem.
Wire's dead page threshold isn't a fixed number. It's relative: `max(10, median * 0.05)`. A page on a niche B2B site with 12 impressions might be healthy. The same page on a broad site is dead. The cannibalization thresholds work the same way: ratio alone isn't enough, so Wire requires both a ratio above 0.4 and a skew above 0.7 before flagging a hard conflict. A page can have significant keyword overlap and still not be cannibalizing if traffic is evenly split. Which situation are you in?
When Wire merges two pages, it picks the keeper by score: 40% impressions, 30% position, 20% clicks, 10% keyword count. The losing page's content gets absorbed, not deleted. Its `index.md` becomes `index.md.archived`. The weighting means a page with strong demand but a weak position can still win over a page that ranks well for low-volume terms. That surprises people. The question is whether the score reflects what you'd choose manually, and if not, which signal is pulling the result in the wrong direction.
Google rewrites 76% of title tags now, up from 61% two years ago. The lowest rewrite rate, around 40%, happens in the 51-60 character range. Go over 70 characters and the rewrite rate hits 99.9%. Wire targets 51-55 characters and strips brackets and pipe separators, both patterns that trigger rewrites. But title rewriting is only one shift. AI Overviews are now present on enough queries that organic CTR on informational searches dropped 61% between mid-2024 and late 2025. A page holding position 3 with 1% CTR may be performing well by current standards, not underperforming.
Position 1 used to deliver around 28% CTR. GrowthSrc measured it at 19% in 2025, and that's the average across all query types. For niche B2B queries where AI Overviews answer the question directly, position 1 delivers 2-3%. Positions 6-10 actually increased 30% as users scroll past AI-summarized results. This is why Wire scores opportunities using `impressions * (1 - CTR)` rather than position alone. High impressions with low CTR signals demand that isn't being captured, which is a different problem than low rankings. The benchmarks you're using to evaluate performance may be three years out of date.
The December 2025 core update shifted Google's evaluation from "content quality" to "content necessity." Generic rewritten content lost ground. Original data, case studies, and first-hand experience gained. The August 2025 spam update moved fast, with near-immediate impact on scaled content and expired domain abuse. The February 2026 Discover update was the first dedicated to that surface, reducing clickbait and favoring locally relevant, in-depth content. If your traffic dropped in a specific window, the update table maps dates to likely causes. The March 2026 core update was still rolling out when this page was written.
NavBoost uses 13 months of rolling click data and is described in the leak as one of the strongest ranking signals. Wire aligns with this by tracking GSC click and impression patterns over time rather than treating each crawl as a snapshot. Internal linking gets flagged when inbound link counts fall below 3, based on SearchPilot's orphan page experiments. H1 keyword alignment with the title showed a 28% traffic increase in controlled tests. These aren't correlations from a survey. They're from split tests on live sites. The question is whether Wire is tracking the signals that explain your specific result.
Wire ignores readability scores, keyword density, image alt text for web rankings, Core Web Vitals as a growth lever, schema markup as a direct ranking factor, and canonical tags on static sites. Each one has a reason. Portent analyzed 750,000 pages and found zero correlation between readability and rankings. The API leak confirmed NavBoost shifted weight to user behavior, making keyword density irrelevant. Core Web Vitals exist as a gate, not a growth signal. If you've been optimizing for any of these, the effort wasn't hurting you. It also wasn't moving rankings. The signals that move rankings are elsewhere.
Wire's SEO decisions use specific numbers. Every threshold has a source. Google's public statements are treated as unreliable. The 2024 API leak proved Google uses signals they publicly denied: click data via NavBoost, domain authority, Chrome browsing data. Evidence is ranked: leak-confirmed > A/B tested (SearchPilot) > correlation study (Ahrefs, Zyppy) > case study > expert consensus.
Reading the Audit Output
The audit command produces four sections:
HEALTH. Pass/fail indicators using + (clean) and - (problem found). Each line covers one check category. A clean site shows all + marks.
ACTION. Problems that need fixing, with the exact command to fix them. Only appears when HEALTH shows - indicators.
SEO. Reword opportunities ranked by score, plus content gaps where no page owns a keyword cluster. These are opportunities, not problems.
INFO. Summary statistics: total pages, archived pages, pages without GSC data.
The 2026 Search Landscape
Three shifts define SEO in 2026. Wire's thresholds account for all of them.
AI Overviews Are Destroying Organic CTR
Seer Interactive studied 3,119 informational queries across 42 organizations (25.1M organic impressions, June 2024 to September 2025). Organic CTR on queries with AI Overviews dropped from 1.76% to 0.61%, a 61% decline. Even queries without AI Overviews saw CTR fall 41%. Paid CTR dropped 68%.
Ahrefs confirmed this in December 2025 with 300,000 keywords: AI Overviews reduce position-1 CTR by 58%, up from 34.5% in April 2025. The impact nearly doubled in eight months.
GrowthSrc tracked 200,000+ keywords and found position 1 CTR dropped from 28% to 19% overall. Positions 6-10 CTR increased 30.63% as users scroll past AI-summarized top results.
Zero-click searches grew from 56% to 69% between May 2024 and May 2025 (Similarweb). AIO queries hit 83% zero-click. Google's experimental AI Mode reaches 93%.
Wire's response: impressions-based opportunity scoring (impressions x (1 - CTR)) weights demand over clicks. This matters because a page at position 3 with 1% CTR is performing well by 2026 standards, not underperforming.
Sources: Seer Interactive, September 2025, Ahrefs, December 2025, GrowthSrc, 2025, Similarweb, May 2025.
Google Rewrites 76% of Title Tags
Zyppy's Q1 2025 study found Google now rewrites 76% of title tags, up from 61% two years prior. When rewriting, Google removes ~2.71 words and retains only 35% of the original.
The sweet spot: titles in the 51-60 character range get rewritten ~40% of the time (the lowest rate). Over 60 characters: 76%+ rewrite rate. Over 70 characters: 99.9%.
Wire's response: title length target is 51-55 characters. The _sanitize_content() auto-fix enforces dash separators (not pipes) and strips brackets, both patterns that trigger rewrites. The lint system flags titles outside the range.
Source: Zyppy, Q1 2025.
AI Overview Citations Are Decoupling From Organic Rank
Ahrefs studied 863,000 keywords with 4 million AI Overview URLs in February 2026. Only 38% of AIO citations come from top-10 organic pages, down from 76% in July 2025. 31% of citations come from pages not in the top 100 at all.
Google self-cites 17.42% of all AI Mode citations (SE Ranking, February 2026), up from 5.7% in June 2025. Including YouTube, Google-controlled properties account for ~20% of all sources.
Wire's response: Wire focuses on content quality and topical authority rather than pure position chasing. The enrich pipeline builds pages that answer queries thoroughly, which is what AIO citation selection appears to reward.
Sources: Ahrefs, February 2026, SE Ranking, February 2026.
CTR Benchmarks by Position (2026)
Traditional benchmarks are from before AI Overviews, ChatGPT Search, and Perplexity started intercepting traffic. Here are both the traditional and measured 2026 numbers:
| Position | Traditional CTR | With AI Overviews (2026) | Niche B2B estimate |
|---|---|---|---|
| 1 | ~28% | ~19% (GrowthSrc) | ~2-3% |
| 2 | ~15% | ~12.6% (GrowthSrc) | ~1.5% |
| 3 | ~10% | ~8% est. | ~1% |
| 4-5 | ~5-7% | ~4-5% est. | ~0.5-0.7% |
| 6-10 | ~2-4% | ~3-5% (30% increase) | ~0.3-0.5% |
The niche B2B column reflects what small sites actually see. A position 1 result for "invoice processing benchmark" delivers modest clicks when AI Overviews answer the query directly. Impressions remain a stronger signal than clicks for these sites.
Sources: First Page Sage, 2026, GrowthSrc, 2025.
Recent Algorithm Updates
| Update | Dates | Impact |
|---|---|---|
| March 2025 Core | March 13-27 | Standard core update |
| June 2025 Core | June 30 - July 17 | Standard core update |
| August 2025 Spam | August 26 - September 22 | Targeted scaled content abuse, expired domain abuse, site reputation abuse. Near-immediate impact. |
| December 2025 Core | December 11-29 | Shifted evaluation from "content quality" to "content necessity." Original data and real experience now outperform rewritten generic content. |
| February 2026 Discover Core | Late February | First-ever dedicated Discover update. Reduces clickbait, favors locally relevant and in-depth content. |
| March 2026 Core | Early March, rolling out | ~2 weeks expected |
The December 2025 update is the most relevant for Wire users. It rewards pages with original research, case studies, and first-hand data. Wire's news pipeline and source citation requirements directly support this signal.
Sources: Search Engine Journal algorithm history, Google Search Status Dashboard.
SearchPilot A/B Test Results (2025)
Controlled split tests on live sites. These are the closest thing to causal evidence in SEO.
| Test | Result | Confidence |
|---|---|---|
| H1 keyword alignment with title | +28% traffic | High |
| Pros/cons table on comparison pages | +50% traffic | High |
| Question-based title phrasing | >5% organic sessions uplift | 95% |
| Title with city/clinic at start | +8.5% traffic, ~0.7% CTR increase | 95% |
| Removing embedded maps from location pages | -7% traffic | Significant |
| Airport codes in flight titles (jargon) | -16% traffic | 95% |
Takeaway for Wire users: structure your pages with comparison tables and question-based headings. Avoid jargon in titles. Wire's section structure enforcement (via topic _create.md prompts) makes these patterns repeatable.
Source: SearchPilot case studies.
Threshold Reference
Dead Page Detection
| Parameter | Value | Rationale |
|---|---|---|
| Impression floor | max(10, median * 0.05) |
Relative to site size. Adapts to niche vs broad. |
| Minimum page age | 180 days | New pages need time to accumulate signals. |
Evidence: HigherVisibility recommends fewer than 100 impressions over 3-6 months as a pruning trigger. HubSpot pruned 3,000 posts and saw 106% increase in organic views. Wire's relative threshold prevents pruning healthy pages on low-traffic niche sites.
Keyword Cannibalization
| Parameter | Value | Rationale |
|---|---|---|
| Minimum shared keywords | 3 | Avoids noise from coincidental single-keyword overlap. |
| Hard overlap ratio | > 0.4 | 40%+ of the smaller page's keyword identity is duplicated. |
| Hard overlap skew | > 0.7 | One page dominates 70%+ of shared-keyword traffic. |
| Soft overlap ratio | > 0.15 | Low bar catches "Google confused" pairs (50/50 traffic split). |
| Minimum shared impressions | 100 | Below this, overlap is statistical noise. |
Evidence: TheMetaBlog recommends 100 impressions minimum before analyzing cannibalization. First Page Digital uses 30% shared impressions as the threshold. Wire's dual signal (ratio + skew) classifies overlap more precisely than ratio alone.
Opportunity Scoring
| Parameter | Value | Rationale |
|---|---|---|
| Position floor | > 3 | Top 3 already performing well by 2026 standards. |
| Position ceiling | <= 30 | Beyond page 3, content alone will not move the needle. |
| Minimum score | 15 | Filters low-volume noise. |
| Formula | impressions * (1 - CTR) |
High demand + low capture = opportunity. |
Content Quality
| Parameter | Value | Rationale |
|---|---|---|
| Title length | 51-55 chars | Zyppy: lowest rewrite rate (~40%) at 51-60 chars. |
| Description length | < 160 chars | Google truncation point. |
| Thin content | < 200 words | Floor for flagging stubs, not a target. |
| Source concentration | > 3 links + > 20%, or > 40% + >= 5 total | Dual threshold avoids false positives. |
| Underlinked threshold | < 3 inbound links | SearchPilot orphan page experiment baseline. |
SEO Reword Tiers
| Tier | Coverage | Treatment |
|---|---|---|
| Full | Top 20% | Complete SEO rewrite: headings, body, title, description. |
| Light | Next 30% | Title and description only. High CTR impact, low cost. |
| Skip | Bottom 50% | No action. Low opportunity or already ranking well. |
News Freshness
| Topic type | Default interval | Rationale |
|---|---|---|
| Vendors | 21 days | Fast-moving: M&A, partnerships, product launches. |
| Capabilities | 60 days | Reference content, changes slowly. |
| Guides | 120 days | Evergreen, rarely needs updates. |
Evidence: Siege Media found the average page-one content was updated every 2 years. Wire's intervals are more aggressive because the news pipeline is automated. The cost of checking ($0) is far below the cost of stale content.
Signals Wire Tracks
| Signal | Method | Evidence |
|---|---|---|
| Click/dwell behavior | GSC data + NavBoost alignment | Leak-confirmed: NavBoost uses 13-month rolling click data, "one of the strongest" ranking signals. |
| Title optimization | 51-55 chars, dash separators, no brackets | Zyppy Q1 2025: 76% rewrite rate, 51-60 chars = lowest at ~40%. |
| Heading structure | H1 validation, hierarchy checks, keyword alignment | SearchPilot: H1 keyword alignment = +28% traffic. |
| Internal linking | Inbound count, broken link detection, crosslinking | SearchPilot: orphan pages underperform. Leak confirms link graph signals. |
| Source diversity | Domain concentration analysis | Reboot Online: pages with outbound links to authoritative sources rank higher. |
| Content freshness | Date tracking, news integration, staleness detection | December 2025 core update rewards original, current content. |
Signals Wire Deliberately Ignores
| Signal | Why ignored | Evidence |
|---|---|---|
| Readability/Flesch score | Zero correlation with rankings. Clarity matters; grade level does not. | Portent 750K pages. Ahrefs 15K keywords. |
| Keyword density | NavBoost shifted ranking weight to user behavior signals. Topical coverage matters, not word repetition. | 2024 Google API leak. |
| Image alt text | Ranking factor for Google Images only, not web search. Accessibility benefit, not SEO lever. | Moz study. Leak mentions imageQuality but no alt-text ranking connection. |
| Core Web Vitals | Weak correlation with rankings. A gate, not a growth lever. Theme/hosting concern, not content. | Perficient/Eric Enge study. Leak confirms it exists but weighted far below content/links/clicks. |
| Schema/structured data | No direct ranking change. Feeds entity signals indirectly via CTR. Theme responsibility. | SearchPilot A/B: no ranking change. |
| Canonical tags | No evidence of ranking issues on static sites with stable URLs. | Build system handles automatically. |
Keeper Score (Merge Decisions)
When merging overlapping pages, Wire selects the keeper using a weighted score:
score = 0.4 * norm(impressions) # Demand signal
+ 0.3 * norm(position) # Ranking strength
+ 0.2 * norm(clicks) # User preference
+ 0.1 * norm(keyword_count) # Topical breadth
Each component is normalized to 0-1 across all pages in the topic. The page with the higher score keeps its URL. The other page's unique content is absorbed and its index.md is renamed to index.md.archived.
GSC Changes (Late 2025 - Early 2026)
Two relevant additions to Google Search Console:
Branded vs. non-branded query filter (November 2025). Uses AI-assisted classification including brand name in all languages, typos, and unique products. Wire's keyword analysis already separates brand queries from organic opportunities through impression ratio scoring.
AI-powered natural language configuration (February 2026). Performance reports now accept natural language queries. This does not affect Wire's API integration, which reads raw data directly.
Source: Google Search Central Blog, November 2025.
Sources
All evidence used in Wire's threshold decisions:
- Zyppy. 81K title tag study (Q1 2025), 23M link analysis. zyppy.com
- SearchPilot. Controlled A/B split tests on live sites. searchpilot.com
- Seer Interactive. 3,119 queries, 42 organizations, AIO CTR impact. seerinteractive.com
- Ahrefs. 300K keyword AIO study (December 2025), 863K keyword citation study (February 2026). ahrefs.com
- GrowthSrc. 200K+ keyword CTR study. growthsrc.com
- Backlinko. 15K keyword correlation study, SEO pricing survey. backlinko.com
- Portent. 750K page readability analysis. Zero correlation with rankings.
- HubSpot. Pruned 3,000 posts, 106% organic traffic increase.
- Google API Leak (2024). Confirmed NavBoost, domain authority, Chrome data signals.
- Perficient/Eric Enge. Core Web Vitals: weak correlation, "gate not growth lever."
- Reboot Online. Outbound links correlation experiment.
- SE Ranking. Google self-citation study, February 2026. seranking.com