Short answer: this page lists the 29 named findings that came out of Blogging Titan’s own research between June and October 2026. Each one has a plain definition, the number it rests on, the date we measured it and a link to the study with the full method. If you quote a term, link to the study it came from.
We name a finding when the same pattern keeps turning up in our data and there is no existing word for it. A name makes a result easier to remember, easier to argue with and easier to trace back to the evidence. Every term below was checked with an exact-phrase Google search before we used it, to make sure we were not borrowing someone else’s word for something different.
The full list of our studies is ordered by date. This page is ordered by idea. For general vocabulary (GEO, AEO, retrieval, grounding) use our AI search glossary.
How AI search picks its sources
Eleven terms from our studies of Google AI Overviews, Google AI Mode, ChatGPT and Perplexity. Read together they say one thing: which sources an AI answer cites depends heavily on how the question is asked and when, and much less on the page features most advice tells you to add.
The citation lottery
The volatility in AI Overview source lists on a minority of queries: the same search, repeated by the same person within an hour, returns a partly different set of cited sources. Number behind it: 17 of 20 AI Overviews returned an identical source list on every pull; on the other 3, only 30 to 40 percent of citations survived all three pulls. Measured July 2026. Read the citation lottery study. An 87-day re-pull found 74 of 212 July citations still in place (34.9 percent).
The paraphrase re-roll
Rewording a question, not repeating it, is what reshuffles the sources an AI Overview cites. Number behind it: 14.0 percent of cited sources (37 of 265 question-source pairs) appeared for all three phrasings of the same question; 0 of 19 questions returned an identical list across phrasings. Measured 31 August 2026. Read the paraphrase study.
The asker effect
AI answers hold fairly steady within the hour but change with who is asking and how they word it. The term joins the two findings above and frames our expert roundup on getting cited (October 2026).
The how-to blackout
ChatGPT’s habit of answering beginner questions from its own training without opening the web, so nobody gets cited. Number behind it: ChatGPT cited no source on 48 of 100 blogging questions; it searched the web for 15 of 15 tool questions and 1 of 15 getting-started questions. Perplexity cited sources on all 100. Measured 2 September 2026. Read the AI citation league table.
The nameless citation
A page an AI engine quotes that credits no author anywhere: no byline, no author tag, no author in the structured data. Number behind it: 195 of 951 pages cited by ChatGPT and Perplexity (20.5 percent). Read 24 September 2026. Read the authorship study.
The citation age gap
The difference in how far back each engine reaches for its sources. Number behind it: 33.3 percent of ChatGPT’s dated non-documentation sources were over five years old, against 20.1 percent of Perplexity’s. Same 951-page corpus. Read the age study.
The dateless third
The share of AI Overview citations that carry no publication date at all, which means freshness advice cannot apply to them. Number behind it: 33.3 percent of 243 AI Overview citations had no date in schema, meta tags or visible text, rising to 55.7 percent on transactional queries. Mostly product pages and videos, almost never blog posts. July 2026. Read the dateless third study.
The staleness penalty
Freshness earns a page almost nothing in AI Overview citations, but obvious neglect quietly costs it. Number behind it: 156 pages compared, July 2026. Read the freshness study.
The schema plateau
Structured data is now so common that having it no longer separates cited pages from uncited ones. Number behind it: 138 pages checked, July 2026. Read the schema study.
The citation understudy
The source an AI answer quotes in place of the one that ranks for the query. Number behind it: Reddit ranked organically on 30 of 30 blogging queries and Google’s AI Overview named it on 18; on 7 of those 18, the quoted thread was not the thread that ranked. 20 September 2026. Read the Reddit citations study.
The gatekeeper split
When Google AI Mode recommends a UK business, trades answers lean on a review platform while professional services answers lean on the firm’s own website and Google profile. 160 AI Mode answers, October 2026. Read the UK recommendations study.
What AI crawlers actually do on a small blog
Three terms from the server log we have kept on bloggingtitan.com since 10 August 2026. Every request is logged by user agent, so these numbers come from our own server, not a vendor dashboard.
The GPTBot decoy
Blocking GPTBot feels like shutting ChatGPT out, but GPTBot is the training crawler and barely visits. The retrieval agent that fetches a page for a live user, ChatGPT-User, is a different bot. Number behind it: GPTBot made 1 request in 37 days; ChatGPT-User made 1,387. 10 August to 15 September 2026. Read the crawler log study.
The unmoved fetch
Cloudflare’s 15 September 2026 change decided which sites ChatGPT could fetch. On a site outside Cloudflare, it did not change how much ChatGPT fetched. Number behind it: retrieval-crawler requests averaged 43.0 a day before and 43.3 after. Read the Cloudflare change study.
The one-buyer blog
For a small site, a pay-per-crawl scheme is worth whatever one AI company decides to pay, because one agent does almost all the fetching. Number behind it: ChatGPT-User made 89.5 percent of 2,615 retrieval fetches over 60 days to 9 October 2026. Read the pay-per-use analysis.
Who writes page one, and can you trust it?
Seven terms from our audits of the pages that already rank for blogging and money questions.
Source density
The share of a page’s statistics that have a source link beside them. Number behind it: 71.2 percent of statistical claims on 161 page-one blogging pages had no source near them (410 of 576). 1 September 2026. Read the page-one census.
Stat laundering
A number keeps its precision but loses its provenance: each citation strips a caveat until the figure travels alone. Number behind it: seven published claims that AI traffic converts better than search, ranging from 1.31x to 23x, each traced back to its source. August 2026. Read the stat laundering study.
Shovel content
How-to advice published by a company that sells the tools it recommends. The tell is a guide that spends more words on setup than on writing, because setup is where the checkout is. Number behind it: three of the seven page-one guides for starting a gaming blog were ads for the publisher’s own product. September 2026. Read the gaming blog guide.
The Reddit tax
The page-one space that forums and other user content take from independent publishers on blogging searches. Number behind it: Reddit held position 1 on 19 of 30 blogging queries in July 2026. Read the Reddit tax census.
The one-sided page one
A results page where every article has a financial stake and every stake points the same way. Number behind it: seven articles on page one for three Substack vs WordPress searches, seven stakes, all pointing away from Substack. October 2026. Read Substack vs WordPress.
The imported page one
A national result set that is mostly published elsewhere, priced elsewhere and written for somewhere else. Number behind it: 24 of 174 page-one pages for UK blogging questions sat on a .uk domain. September 2026. Read State of UK Blogging 2026.
The regulated page one
For UK money questions, Google fills page one with regulated firms and public bodies, leaving little room for independent blogs. Number behind it: independent blogs held 2 of 79 page-one results for ten UK money questions. October 2026. Read the finance blog guide.
What tools and services really cost
Four terms from our dated price logs, where every figure was read on the vendor’s own page on a stated day.
The unbuyable month
A monthly price a vendor advertises that you cannot buy by the month, because the smallest purchase is an annual or multi-year prepayment. Number behind it: 9 of 30 blogging tools. 3 September 2026. Read the Blogging Tool Cost Index.
The times-twelve gap
The difference between an advertised monthly price multiplied by twelve and what a year without a lock-in really costs. Number behind it: for 13 of 16 AI SEO and blogging tools, the monthly price times twelve understated the year. 30 September 2026. Read the true annual cost study.
The tracked-question price
The monthly cost of watching one prompt in a paid AI visibility tracker: the plan price divided by the prompts it tracks. Number behind it: across eight trackers with a published price and prompt allowance, it ran from $0.67 to $3.96 a month, a near six-fold gap. 4 November 2026. Read the AI visibility tracker price comparison.
The price-in-writing ladder
A five-rung scale, from no published price to a complete price with terms, for comparing how openly different markets price their services. Number behind it: 5 of 39 UK agencies that publish prices stated their setup position and passed the leave-date test. October 2026. Read the UK marketing prices study.
The dollar-quote gap
In our test, which product a UK business asked an AI about mattered more than which AI it asked. Number behind it: 6 of 35 answers were right on four software vendors, against 69 of 80 on the other software questions. The four were chosen because the engines struggled with them, so part of the gap is built into the selection. October 2026. Read the price accuracy study.
How niche blogs actually make money
Two terms from reading the ads.txt files of the independent sites that rank in each niche. An ads.txt file lists the ad companies allowed to sell a site’s ad space, so it shows who manages a site’s ads.
The managed-ads default
On recipe searches, the independent sites that win have almost all handed their ad space to a managed network. Number behind it: 25 of 25 readable ads.txt files on independent recipe sites named Raptive or Mediavine (Raptive 21). October 2026. Read the food blogger study.
The gaming network mismatch
The ad networks built for gaming sites are the ones a gaming blog is least likely to qualify for, so small gaming blogs use general blogger networks. Number behind it: none of the 14 independent gaming sites on page one named a gaming-specialist network in their ads.txt. 29 October 2026. Read the gaming blog monetisation guide.
The niche commission gap
How far a blog’s niche moves the Amazon Associates rate it earns. Number behind it: on Amazon.co.uk, £1,000 of referred sales pays £50 in a food or personal finance niche and £10 in a gaming niche, a five-fold gap. 4 November 2026. Read the Amazon affiliate program guide.
How to cite these terms
Every term above is free to use. Please link to the study it came from, not only to this page, so readers can check the method. A suggested form: “the paraphrase re-roll” (Blogging Titan, 2026), linked to the study URL. When we re-run a study, the earlier edition stays online with its date, so a link you add today will keep pointing at the numbers you quoted.
Found one of these terms used without a link back, or think one of our findings is wrong? Tell us through the contact page. Corrections are logged on the study itself. Our testing method and editorial policy explain how the studies are run.