Original data · SEO Tools
Search Console clicks for October 2026 in each site's main country (Bulgarian, Slovak, Romanian, UK and U.S. search; 101,141 clicks) against a keyword-based estimator for the same country: the median site was estimated at 35% of its real clicks, 9 of 15 at under half and 2 at under a tenth. The keywords the tool knew were overestimated 2.4 times, while 81% of real clicks came from searches it did not list (40% anonymized queries). Every "accurate" total was two errors cancelling out. Compared with Ahrefs and Screaming Frog studies, with a 6-step check and a correction-factor calculator. Downloadable data, CC BY 4.0.
Published November 2, 2026 · 12 min read
Guide · Hiring SEO Help
Who to hire for which problem, where to find candidates (5 of the top 10 US results for "hire seo consultant" are marketplaces), a one-page brief, eight questions with good and bad answers, a paid audit or trial, contract and access terms, red flags and what to ask about AI search - built on Google's own advice on hiring an SEO.
Published October 28, 2026 · 13 min read
Original data · AI Search
407 mobile queries in Bulgarian, Slovak and English search, checked on October 26, 2026: 65% of the pages Google's AI Overviews cited did not rank in Google's top 100 for the same query, and only 28% were in the top 10. For ChatGPT, 91% of 932 cited pages were outside the top 100 and only 9% overlapped with the AI Overview. Ranking still helps: position 1 was cited 34% of the time, position 11 just 7%. Where non-ranking citations come from, how ChatGPT rewrites the search, how much citations change in a week, and a 5-step check. Downloadable data, CC BY 4.0.
Published October 26, 2026 · 13 min read
Original data · SEO
Search Console data from 15 sites and 2.85 million impressions: the median site-wide CTR was 4.1%, but only 1.2% on non-branded queries, because branded searches brought 24% of clicks from 3.5% of impressions. On the median site, positions 4-10 held 58% of non-branded impressions at a 1.4% CTR, while positions 1-3 held 19% and earned 37% of clicks. Dropping page 2 lifted non-branded CTR by just 17% on the median site, and 43% of impressions never appear in the query table. Benchmarks from Ahrefs and Databox, a year-over-year check, a 5-step diagnosis and a calculator. Downloadable data, CC BY 4.0.
Published October 26, 2026 · 13 min read
Original data · AI Search
Search Console data from 15 sites, mostly in Bulgarian search: on the same query at a stable position, mobile CTR fell 28% from autumn 2025 to autumn 2026 where Google now shows an AI Overview, against a statistically unclear -10% where it does not. Impressions rose 54% while clicks fell 12%, top-3 positions lost 35%, and being cited inside the overview did not save the click. AI Overviews showed on 71% of page-one queries (74% Bulgarian, 55% Slovak), but on only 19% of queries with a map pack. Why the common with-vs-without split points the wrong way, a step-by-step method and a calculator. Downloadable data, CC BY 4.0.
Published October 19, 2026 · 13 min read
Buyer's guide · AI Search
Google's AI Overviews cited self-promotional "best" lists 323 times in a 100-query test and did not recommend the author in 224 of them (69%). How often AI answers cite pages a brand controls, why that rarely becomes a recommendation, what Google does with self-promotion, US FTC 16 CFR 465.6 and EU and UK review law (up to $53,088 per violation in the US), research on manipulating AI search, whether paid awards help, an 8-question risk checklist, and seven popular claims that do not hold as stated. Updated with July to October 2026 data: self-promotional lists fell from 44% to 24% of the listicles AI Overviews cite. 37 sources, vendors flagged.
Updated October 19, 2026 · 15 min read
Buyer's guide · AI Search
One tracker saw Reddit's share of ChatGPT Search citations fall from 3.83% to 0.52% in August 2026, while Google AI Overviews fell 11% and AI Mode 31%. Which engine a Reddit agency can still move, four large swings in sixteen months, published prices of $2,500 to $12,500 a month, who owns the accounts, FTC disclosure, how to check citation claims yourself, a 10-question pre-hire checklist, and six popular claims about Reddit and AI that do not hold as stated. 33 sources, vendors flagged.
Published October 5, 2026 · 14 min read
Data report · AI Search
Original experiment: 144 answers from Perplexity, GPT-4o and Gemini to "X vs Y: which is better?" for 12 software vendor pairs. 76% cited a page written by one of the two vendors, yet when a vendor's own page was cited it won 68 times and lost 66, and 94 of 96 clear verdicts went to the same winner per pair. What a competitor's page really moves is the facts, with a step-by-step fix for false claims (fact page, third-party repetition, NAD, Lanham Act, EU Directive 2006/114/EC) and for answers dominated by one negative review. Raw data published.
Published September 28, 2026 · 11 min read
Data report · AI Search
Analytics puts AI assistants under 2% of tracked leads - 0.073% of 20 million inbound calls (CallRail). Companies that ask their customers report 7% to 25% (Outfunnel, Webflow, Vercel, Tally), and 45% to 94% of buyers say they research with AI. The gap has a mechanism: 55.9% of visits after a ChatGPT recommendation arrive as a branded search. Three layers side by side, a gap calculator, and four popular figures that did not trace to their credited source.
Published September 21, 2026 · 10 min read
Data report · AI Search
Original pilot: 176 citations logged and every cited page labeled by shape and owner. "best X" questions cited listicles 84% of the time and comparison pages 2%; "X alternatives" flipped to 86% comparison pages, a quarter of them on competing vendors' own domains, while G2 earned one citation. Why the published "under 3%" figure for comparison pages is an average that hides the switch. Raw data published.
Published September 7, 2026 · 9 min read
Data report · AI Search
Aggregated, source-verified 2026 data: AI-referred visitors convert about 50% higher than organic on Shopify product pages and are worth 4.4x per visitor (Semrush), with B2B lead conversion of 14.2% vs 2.8% for Google organic (Opollo). The catch the hype skips - volume is still tiny, and AI converted 43% worse than other channels as recently as July 2024. What it means for a small business, with limitations named.
Published August 31, 2026 · 9 min read
Data report · Local Search
Original dataset: 7,072 Google Business Profile listings across 6 home service trades in 12 US cities. The most-reviewed company held a top-3 Maps position in just 5.6% of 72 markets, the top 5 listings hold a median 58.3% of visible review volume, and trade predicts difficulty far better than city size. Full method, limitations and raw data published.
Published August 24, 2026 · 9 min read
Data report · AI Search
I ran the same four buyer questions 60 times across ChatGPT, Gemini and Perplexity. Which engine you track moved a brand's measured visibility 2.3x more than the model's own randomness - and two engines recommended a tool the third never named once. Original data, full method and limitations, raw dataset published.
Published August 17, 2026 · 8 min read
Guide · Hiring SEO Help
A working SEO professional's guide for US businesses: what the job includes in 2026 (including AI search), when DIY is enough, engagement models, what business owners really say about hiring SEO help, and the red flags to walk away from.
Updated July 30, 2026 · 9 min read
Data report · AI Search
93% of businesses have at least one fact wrong or missing in AI answers. Aggregated, source-verified statistics on how ChatGPT, Gemini and Perplexity misstate business hours, phones and services - with a 10-prompt self-audit protocol to test your own business.
Updated July 20, 2026 · 9 min read