The Citation Floor Method
AI engines do not rank you. They cite you. Citation is not a score you climb, it is three floors you either clear or you do not. Clear all three and your page becomes eligible to be cited. Stay below any one and you stay invisible, however much you spend elsewhere.
The Citation Floor Method is how GeoLinks earns AI-search citations for a brand. It treats citation as three thresholds rather than a ranking score: a page must be extractable, it must be fresh, and the brand must hold enough off-site authority. We diagnose which floor a site is stuck on, raise that floor, then measure Citation Share of Voice across ChatGPT, Claude, Gemini, Copilot, Grok, Perplexity, Google AI Mode and AI Overviews. The diagnosis decides the work, not a fixed package.
Diagnose before you spend
You cannot clear a floor you have not measured. Every campaign opens with a baseline across all five engines: prompt research to find the queries your buyers actually use, the free AI Visibility Check to see who gets cited for them today, and a backlink audit to place your authority against the trust threshold. The diagnosis tells us which floor is lowest. That floor gets the next pound of budget.
Clear all three, or stay invisible
The lowest uncleared floor is what holds the whole programme back. A perfectly structured page with no off-site authority still does not get cited. A heavily linked page with an unquotable layout still does not get cited.
Extractability
The page has to be quotable
AI engines cite passages they can lift cleanly. Open every commercial page with a 50 to 75 word answer block under the H1, add FAQPage schema, and pair text with a table or image. If a page makes the engine work to find the answer, it quotes whoever wrote a cleaner one instead.
What the data shows
FAQ schema carries around 40% of ChatGPT source-selection weight. Multimodal pages correlate 0.92 with AI Overview selection, the highest single factor measured in 2026.
The honest part
General product and Organization JSON-LD does not move citations. Ahrefs' May 2026 causal test across 1,885 pages found no uplift. We add schema for rich results and entity clarity, not for citations.
Freshness
Stale pages get skipped
Recency is a ranking input the AI layer weights heavily. A page touched inside the last 30 days is treated as more reliable than one last edited in 2024. Freshness is not a launch task, it is a maintenance floor you hold every month with real updates: a new statistic, an updated table, a revised answer.
What the data shows
Content updated within 30 days earns 3.2x the ChatGPT citation rate of content older than 90 days, measured across our client campaigns.
The honest part
A date change with no real edit does nothing. The engines sample the content, not the timestamp. Cosmetic freshness is theatre.
Authority
Trust is earned off your own site
Around 91% of AI citations point at third-party sources, not at the brand's own pages. You clear this floor off-site: niche edits into pages the engines already cite, guest posts on vetted sites with real traffic, and a verified review-platform presence. This is the Links half of GeoLinks, and it is where most programmes stall.
What the data shows
Citation consistency jumps sharply above roughly 32,000 referring domains. A verified Trustpilot, G2 or Capterra profile carries about 3x the citation probability of having none.
The honest part
Domain Rating alone no longer predicts citations. Its correlation fell from 0.55 in 2024 to 0.18 in 2026. Volume of relevant referring domains still gates trust, but a high DR with thin topical links will not carry you.
Measure, then fix the lowest floor again
Citation Share of Voice is the number that matters: how often you are cited versus competitors for the prompts your buyers use. We track it monthly across all five engines, then return to the diagnosis and raise whichever floor is now lowest. AI search shifts fast, so the method is a loop, not a one-off project.
ChatGPT
Claude
Gemini
Perplexity
Overviews
AI search referral share, Goodie, 21 May 2026. The method tracks Citation Share of Voice across all five.
The floors, raised on real sites
Extractability + freshness
monthly organic visits in 7 months, +776%
Garden Ornaments. We raised the extractability and freshness floors with no net new referring domains. When authority was layered on top, the citation rate compounded.
Read the case study →Authority
on a fresh domain in 30 days, 0 to 27 referring domains
Garden UK launched with zero authority. Targeted niche edits and structured content cleared the first authority threshold and earned 10 ranking keywords in the first month.
Read the case study →What the Citation Floor Method is not
Most of what is sold as AI-search optimisation does not clear a single floor. We say so on the record, because pretending otherwise wastes your budget and our reputation.
- ✗ It is not schema theatre. Bolting JSON-LD onto every page does not earn citations. FAQ schema helps extraction; the rest is for rich results and entity clarity.
- ✗ It is not an llms.txt file. No major engine shows a measured citation lift from one today. Ship it as future-proofing, not as a tactic.
- ✗ It is not a monitoring dashboard. A tool that tells you that you are invisible is a diagnosis, not a fix. The work that follows is what changes the number.
- ✗ It is not vanity mentions. Being named in a model's training answer is not the same as being cited as a live source. We count citations, not mentions.
Method FAQ
What is the Citation Floor Method?
It is GeoLinks' framework for getting cited by AI search. AI engines do not rank you, they cite you, and citation depends on clearing three thresholds: extractability, freshness, and authority. Clear all three and a page becomes eligible to be cited. Stay below any one and it stays invisible, regardless of spend elsewhere.
Why call them floors rather than ranking factors?
Because they behave like minimum thresholds, not a score you climb. A page with perfect structure and no off-site authority still does not get cited. A heavily linked page with an unextractable layout still does not get cited. Each floor has to be cleared. The lowest uncleared floor is what holds the whole programme back.
How is this different from your how-it-works process?
How It Works is the operational delivery: content, links, publication, delivery report. The Citation Floor Method is the model behind it. It explains which floor a given site is stuck on and where the next pound of budget should go. The diagnosis decides the work, not a fixed package.
Which floor should a small business fix first?
Usually extractability and freshness, because both are largely free and within your control. Authority is where external budget becomes necessary. Diagnosing first prevents the common mistake of buying links while the pages they point at are still unquotable.
How do you measure whether the method is working?
Citation Share of Voice across ChatGPT, Claude, Gemini, Copilot, Grok, Perplexity, Google AI Mode and AI Overviews, tracked monthly. We re-run the AI Visibility Check, compare the before and after, and identify the lowest remaining floor. The number that matters is how often you are cited versus competitors for the prompts your buyers actually use.
Does the Citation Floor Method work outside ChatGPT?
Yes, and it has to. ChatGPT's share of AI search referrals fell from 89.1% to 62.6% in under a year while Claude reached 18.5%. The three floors apply across every engine. Around 80% of the work is shared, and the remaining engine-specific tuning sits inside the authority and extractability floors.
Find out which floor you are stuck on
The free AI Visibility Check scans all five engines and shows your current citation share. It is the first step of the diagnosis, and it takes under five minutes with no call.