Trust Flow, Citation Flow and Topical Trust Flow

Lesson 13 of 50Module 3 · Reading the link graph6 min read

By Bill Hartzer · Hartzer Consulting · 25+ years in search, domains and expert witness work

What the two numbers actually measure

Citation Flow is a volume measure: how much link citation a page or domain has accumulated, without any judgment about where it came from. Trust Flow is a quality measure: it propagates outward from a manually reviewed seed set of trusted sites, so a page scores well only if there is a short path to it from something known to be good.

Both run 0 to 100 and both are logarithmic, which is the part people forget. The distance from 20 to 30 is not the distance from 60 to 70; the upper bands are enormously harder to reach. That single property invalidates most of the arithmetic done with these numbers — averaging Trust Flow across a link list, for instance, produces a figure with no meaning.

Two calibration points from unmanipulated data. Across 22,260 referring domains on a profile where no link was ever bought, 69.3 percent are Trust Flow 0 and only 446 domains — 2.0 percent — are Trust Flow 41 or above. That is roughly 22 strong domains a year over two decades of publishing. When a vendor offers fifty TF 50-plus links in a quarter, they are offering more than double what a well-known industry site accumulated per year, at higher quality, in a tenth of the time.

The second point: Trust Flow predicts how long a link lasts. On that same profile, the median referring domain stopped linking after 1,080 days. Split by Trust Flow, TF 0 domains had a median life of 859 days and TF 61-plus domains 3,353 days — about four times as long. A strong link is not only worth more while it exists; it exists for longer.

The ratio, and how far to trust it

The relationship between the two numbers is more informative than either alone. Plot Trust Flow against Citation Flow and a healthy profile's referring domains cluster along the diagonal, densest in the low-low corner. A manipulated one sits below it — citation without trust.

Most sites sit slightly below parity and that is unremarkable. The signal is not one domain below the line; it is a population below the line, in the same band, in one tier. That is what the tier-2 and tier-4 clusters in Lesson 3.2 look like numerically.

The way to see this properly is the referring-domain scatter in Compare Sites, which plots every referring domain of one or two sites with Trust Flow on one axis and Citation Flow on the other. Two healthy profiles produce two clouds hugging the diagonal, both densest in the low-low corner, with the stronger site's cloud extending further up and to the right. That corner density is worth staring at for a moment, because it is the honest picture of any profile: most of what links to you is weak, on every site, always. A manufactured profile produces a cloud that sits visibly below the diagonal instead, and you can read that in a second without a single number.

Three cautions I would not skip. Trust Flow above Citation Flow, as on the abandoned domain, usually means a tiny profile where a couple of decent legacy links dominate — it is not a badge of quality. A low ratio on a young site often just means young. And the ratio says nothing whatever about relevance, which Module 2 argues is the metric that matters most and the one nobody sells.

Topical Trust Flow, and the specific way it misleads

Topical Trust Flow splits the trust score by subject category, so you can see not just how trusted a site's links are but what they are trusted about. It is the closest thing to a machine-readable relevance signal in common use, and it is weaker than its presentation implies.

Here are five rows from a real Topics tab, and the reason this section exists:

The topic with the most referring domains has the lowest Topical Trust Flow on the page, and the topic ranked first rests on six domains. Rank order is by TTF value, so a category can top the list on a handful of strong referrers while a category with ninety times as many domains sits at the bottom because those domains are weak.

That produces two opposite errors and I have seen both. Reading the rank as importance tells you an SEO blog is chiefly about internet etiquette. Reading the volume column as relevance tells you it is about travel. Neither is true. Sort by TTF value, treat the domain count as sample size, and read the two columns together or not at all.

Three further limits worth carrying:

• The columns do not sum. A domain can carry several topics — on this profile about 1.63 each — so adding the referring-domain column gives you a number larger than the profile.
• Ties are unordered. Capture the same table twice and rows sharing a TTF value swap places. Position within a tie group carries no information.
• Classification is noisy at the item level. On the same profile the brand anchor billhartzer is classified Society/People while www.billhartzer.com is Business/Automotive. Same site, same brand, incompatible labels.

Use it as a weak signal that sharpens a judgment you formed elsewhere. Never as a verdict.

History: the context that stops you overreacting

Two charts turn a snapshot into a trend, and both are chiefly useful for talking people out of panic.

Flow Metric History plots Trust Flow and Citation Flow per index build, going back about a decade. On my own site Trust Flow sits in a band around the low-to-mid twenties from 2017 through 2024, then climbs to the mid-thirties across 2025 and 2026; Citation Flow hovers near 40 for most of the period. There are also several single-day drops to zero. Those are index-build artifacts, not events — one bad data point is a data point, not a penalty. On the abandoned domain the same chart is nearly flat and low for ten years, with one early Citation Flow spike that decays away and never recovers. That is a link blast that did not stick, drawn.

Backlink History plots referring URLs and referring domains over time, and the trick is to read the two curves against each other. On my site there is a single day near the end of the window with roughly 830 new referring URLs and no corresponding rise in referring domains. One domain fired several hundred URLs — a sitewide placement or a scraper — and it is not link acquisition in any sense worth reporting. Any time the URL curve spikes without the domain curve, the answer is one source, not many.

The abandoned domain's chart is the opposite failure: a ceiling of three referring URLs in a single day and a referring-domain count that never exceeds one. A flatline.

How to use these metrics without being used by them

Five rules I would defend line by line:

• No metric is a purchase criterion. Trust Flow is a property of a domain. You are buying a placement on a page. Lesson 3.1 explains why the two can be unrelated.
• Read distributions, not averages. On my profile, 572 of 1,130 live referring domains sit at TF 1-20 and seven sit at TF 81-100. The average of that is a fiction; the shape of it is the profile.
• Compare within a tool and within an index. Trust Flow against another vendor's authority score is not a comparison, it is a category error.
• Expect Trust Flow 0 in bulk. Seven in ten referring domains on a profile with nothing bought are TF 0. Junk is the default state of having existed online, not evidence of an attack.
• Use the metric to prioritize, then look at the page. Metrics narrow a list of thousands to a list of dozens. Human judgment does the rest, and there is no substitute for it.