'Powered By' and Attribution Links: Product-Led Link Acquisition

TL;DR

• A powered-by link is a pricing artifact, not a link tactic. Your customers distribute it; the only lever you control is the terms under which they can remove it.

• Google’s spam policy on contractually required links carries a qualifier almost every summary drops: the violation is requiring a link without allowing the site owner the choice of qualifying it. That clause defines the compliant model.

• The guidance is directional: Google tells the site receiving boilerplate attribution not to worry, and tells the vendor who controls the link to nofollow it. The obligation is yours, not your customers’.

• Volume and quality move in opposite directions. The free tier that maximises link count concentrates those links on the cohort that disappears fastest.

• At B2B SaaS median churn of 3.5% a month, an attribution link’s half-life is roughly 19 months against about 7.3 years for an editorial link. Your attribution referring-domain count is a churn report with a lag.

• Use the Attribution Control Map below to place your implementation in one of four cells, then engineer toward the only defensible one.

• Measure it as an acquisition loop, not as links. The test: would you still ship it if the link passed nothing? It already does.

The claim this article argues against

Open almost any guide to product-led link building and you find the same sentence in different clothes: widgets, badges and embedded features each carry a branded link home from every site that installs them, making attribution a steady source of passive, homepage-level authority. It is the most repeated claim in the category, and wrong in three checkable ways. The interesting part of this topic starts once you take it apart.

Take the words in order. Homepage-level authority assumes these links pass equity. They are sitewide boilerplate links in a footer or widget chrome, and Google’s published link-spam examples name that shape twice over: keyword-rich links embedded in widgets distributed across various sites, and widely distributed links within footers and templates. Not an edge case that might get caught. The named category.

Passive is true, and that is the problem rather than the benefit. Passive means you did not choose the domains. Every other part of your programme is built on selection: which publishers to pitch, which assets to build, which prospects clear the bar. A link profile you did not select is one you cannot defend, and at scale it is the part of your backlink graph that most resembles an automated scheme — the same anchor, the same target, the same template position, replicated across every domain running your software.

Steady collapses hardest under measurement. These are the least steady links you will ever hold, because their lifespan is not a function of editorial durability. It is a function of your churn rate, and the numbers below are worse than most teams expect.

None of which makes attribution worthless. It means the standard framing has the mechanism backwards, so teams acting on it optimise the one variable — raw link count — that is inversely related to what they actually want.

Attribution is a pricing decision, not a marketing one

Start with the structural difference that makes attribution behave unlike every other link asset. With an interactive tool or calculator you host and promote, you build the asset and pitch a publisher. With a certification badge, you set a standard and award it. Distribution is an act you perform, one relationship at a time, and you can stop.

Attribution is not distributed by you at all. It ships inside the product and appears on a domain because a customer signed up, installed something and stayed on a particular plan. Nobody in marketing approved any individual placement. The link is a downstream consequence of a line in the pricing table.

The organisational consequence is one most SEO teams discover the hard way: you do not own this channel. Whether attribution is required, optional, removable or purchasable is a pricing decision made by product and finance. When a problem surfaces — a bad anchor, a hardcoded followed link, a wave of installs on junk domains — no outreach change or disavow file fixes it. A release does. If you work as a link building specialist inside a software business, your job here is to write the requirement and hand it to the people who can implement it, not to run a tactic.

Four commercial models exist in practice, producing very different link profiles from identical products.

Model	How the link arises	Who controls the rel attribute	Resulting profile
Required on free tier	Condition of the free plan; hardcoded	Vendor, unless a setting exists	Highest volume, lowest quality, fastest decay
On by default, removable	Ships enabled; switchable in settings	Customer, in effect	Moderate volume, self-selecting quality
Paid removal	Required until a branding-removal add-on is bought	Vendor, priced	High volume; most often misread as a scheme
Opt-in advocacy	Off by default; customer chooses to display	Customer	Lowest volume, highest quality, genuine signal

Read that as a menu of trade-offs, not a ranking. Each row is a decision somebody already made, usually years ago, for reasons unrelated to search. Find out which row you are in and who owns it.

What Google actually says — and the clause everyone drops

Google’s spam policies list, among examples of link spam, requiring a link as part of a Terms of Service, contract, or similar arrangement without allowing a third-party content owner the choice of qualifying the outbound link as they deem fit.

Almost every summary quotes that up to the word “arrangement” and stops. The dropped half is the operative half. Google is not prohibiting required attribution. It is prohibiting required attribution the site owner cannot qualify — cannot add rel=”nofollow” or otherwise mark the link as something they did not editorially vouch for.

The single sentence that resolves most of this topic

The violation is not the requirement. It is the requirement plus the removal of choice.

A required link that ships nofollow, or that any customer can qualify from a settings panel, sits outside the clause as written — the choice was never taken away.

Two further examples on the same list govern how the link is built: keyword-rich links embedded in widgets distributed across various sites, and widely distributed links within footers and templates. Those three clauses define the whole compliance surface. Requirement without choice. Keyword-rich anchors. Template-wide distribution of followed links. Hit one and you are in the policy; hit all three and you have built the illustration Google wrote it from.

The asymmetry: whose problem is this?

Here is the under-covered part, and it changes who in your organisation should care. Google’s guidance is explicitly directional — it says different things to the site displaying attribution and to the vendor receiving it.

In February 2023, Google’s Lizzi Sassman addressed sitewide footer links crediting a designer or a CMS. To the host: boilerplate credit arriving with a website theme is not something a site owner needs to worry about. To whoever controls the link, something different — add nofollow, and check the anchor is reasonable rather than gratuitously keyword-rich, her illustration of the failure being a credit reading “made by the best Florida SEO” rather than a brand name. Mueller gave materially the same advice about designer credits in 2015.

On widget links Mueller was blunter in March 2022: they are against the guidelines, trivial for Google’s systems to recognise, and — the line worth pinning above the desk — you can count these links, but that does not mean they have any value. He added that Google may take manual action on unnatural links, and declined to call a broad scheme of this kind harmless.

As an operating rule:

Party	What Google effectively says	What that means operationally
The customer displaying it	Boilerplate credit arriving with software is not their problem	No remediation needed; telling them otherwise is scaremongering
The vendor controlling the link	Nofollow it, and keep the anchor reasonable	The compliance obligation is yours, discharged in a code change
Either party, keyword-rich anchor	The named widget-link failure	Fix the anchor first: cheapest and highest-value change

The upshot is clean: attribution risk needs no campaign or outreach effort. It needs one release. And because disavowing links is a blunt instrument you rarely need, the fix belongs upstream in the component, not downstream in a spreadsheet.

Is charging to remove attribution a link scheme?

The question the model provokes and almost nobody answers directly. Direct answer: not by itself, and the clause above shows exactly where the line sits.

If the link ships nofollow with a branded anchor, a customer paying to remove it is buying white-labelling — the absence of your brand from their interface. No ranking credit was flowing, so none is being sold, suppressed or withheld. Commercially it is identical to charging for a custom domain or a logo swap, one of the oldest pricing levers in software.

It becomes something else the moment the link is followed. Then the free tier accumulates PageRank from sites that never editorially vouched for you, the paid tier is the price of stopping it, and the only route to qualifying the link is a card payment. That is the clause almost verbatim. The money is not the problem; the followed link is, and the money makes the intent legible.

The Attribution Control Map

A scorecard is the wrong instrument here. Attribution is not a judgement call across weighted factors; it is a two-variable system, and both variables come straight out of the policy text — whether the link is a condition of use, and whether the site owner can qualify it. Everything else is detail.

Plot your implementation on those axes and you land in one of four cells.

Cell	Condition of use	Qualification control	Verdict
A — Named Scheme	Required by terms or hardcoded	None: followed, vendor-controlled, often keyword-rich	The policy illustration. Highest exposure, equity neutralised anyway
B — Extractive	Optional or default-on	Followed and hardcoded; no setting exists	Not the named violation, but you collect followed sitewide links nobody chose to give
C — Compliant but unmanaged	Required	Followed by default, but a setting exists	Outside the clause as written, and fragile: it relies on customers exercising a right they do not know they have
D — Defensible distribution	Required or optional	Ships nofollow, branded anchor, visible and removable	The only cell worth engineering toward. Passes nothing by design; value is acquisition

Cell D does not require giving the link away. Attribution can stay mandatory on your free plan and still sit in D, because the clause governs choice of qualification, not choice of display. That distinction is commercially valuable, and it is what teams most often get wrong in the cautious direction — stripping attribution entirely when a rel attribute would have done.

The fastest single tell

Ask internally: “If we shipped this link as nofollow next release, would anyone object?”

If the answer is no, you are in Cell D already or one line of code away from it.

If somebody objects — if a forecast, target or slide depends on those links passing equity — you have been running an equity play and calling it attribution. The objection is the diagnosis.

The volume–quality inversion

Which is why raw attribution counts should never reach a board pack.

Free-tier attribution maximises link count, and the arithmetic is real: tens of thousands of installs generate a referring-domain figure no outreach programme could match at any budget. But free-tier users are your least committed cohort — trials that never converted, side projects, agency test installs, sites for clients who never launched, businesses that closed.

So the domains carrying your attribution skew toward abandoned projects, parked domains, thin sites and staging environments — each carrying your link in a template, on every page. That is the population profile, not the exception within it.

The inversion, stated plainly

Every change that increases attribution link count decreases their average quality, and every change that raises quality cuts the count.

Moving attribution behind a paid tier can cut referring domains by an order of magnitude while improving every remaining one — paying customers are, definitionally, retained customers on real sites with real traffic.

No configuration maximises both. Anyone promising one is counting links, not looking at them.

Elsewhere volume and quality are merely in tension, traded at the margin. Here the pricing decision that produces the links couples them mechanically in opposite directions.

The attribution half-life

Editorial links are durable. Searchlab’s analysis of link persistence puts the average lifespan of an earned backlink at roughly 7.3 years, which is why one strong placement keeps compounding long after the campaign is forgotten.

An attribution link has no editorial durability. Its lifespan is your customer’s lifetime: the account closes, the software is uninstalled, the link goes. So the decay curve of your attribution profile is your retention curve, and you can compute it.

With monthly churn c, the share of an install cohort still live at month t is (1 − c)^t, and the half-life is ln(0.5) / ln(1 − c). Applied to 2026 SaaS retention benchmarks — a B2B median near 3.5% monthly, 0.5–2% for enterprise and annual contracts, 3–7% for SMB self-serve, higher again for free tiers — the picture is this.

Monthly churn	Typical segment	Link half-life	Alive at 12mo	At 24mo	At 36mo
2%	Enterprise / annual contracts	34 months	78%	62%	48%
3.5%	B2B SaaS median	19 months	65%	42%	28%
5%	SMB self-serve	14 months	54%	29%	16%
8%	Free tier / prosumer	8 months	37%	14%	5%

Against 7.3 years the comparison is stark. At the B2B median an attribution link is worth about a quarter of an editorial link in duration alone — before accounting for passing no equity, sitting in a template, and never having been selected. At 8% monthly churn, ninety-five percent of a cohort is gone inside three years.

Where this model is wrong, and which way

It models the customer relationship, not the HTML. Some attribution outlives the account: an abandoned site keeps its stale footer indefinitely.

So real survival runs above the model — inflated by precisely the domains you least want to count on: dead sites, parked domains, pages nobody maintains.

The model is a floor on decay and a ceiling on quality at once. Both readings point the same way, which is why the conclusion holds either way.

The operational consequence is the useful bit: your attribution referring-domain count is a lagging churn report. A 15% fall across two quarters is a retention finding before it is a search finding, and the person who should see it first sits in product. Read alongside your link velocity benchmarks and acquisition rate, it also explains a pattern that confuses many software teams: an attribution-heavy profile can show healthy gross acquisition and flat net growth for years, because install rate and churn rate cancel out.

Reciprocity, leaching, and the anchor problem

Three second-order issues surface once attribution runs at scale. All are more common than the compliance question and get less attention.

Leaching. The agency version — the “designed by” credit in a client’s footer — draws a criticism that applies equally to software vendors: the client paid for the site, and a sitewide link off it was never part of what they bought. Commentators call it double-dipping, and the label lands. It also creates fragility — links acquired without informed consent vanish in a wave once a client base notices them.

Reciprocity. You publish a case study linking out to a customer; their footer links back on every page. One instance is nothing. Across several hundred accounts it is a reciprocal pattern spanning your client base, and reciprocal linking at scale has been a detectable footprint since Penguin. Not a reason to stop writing case studies — a reason not to compound it with a followed footer link.

Anchor uniformity. Every install ships an identical anchor to an identical destination. Even a perfectly branded one produces total uniformity across thousands of domains — the homogeneity link-population analysis exists to detect.

The fix that makes it worse

The instinctive optimisation is to vary the anchor across installs so the profile looks natural.

Do not. Manufacturing variation in a template you control is not diversification — it is footprint evasion, converting a compliance question into an intent question. Randomised anchors across a distributed template are far harder to explain than a uniform branded one.

The honest fix is the boring one: keep the anchor uniform, keep it branded, ship it nofollow, and stop expecting it to be an equity channel.

What attribution actually buys

If the links pass nothing, why ship attribution at all? The answer is not the usual consolation prize about brand awareness. The returns are real, specific and unusually measurable — and none of them live in your backlink tool.

The highest-intent discovery channel in software. Consider what an attribution click is. Somebody has just used a booking form, chat widget or scheduling page — and it worked. They look down, see who made it, and click. Not mid-research comparing vendors on a listicle; mid-experience, having watched the product work on a live site in their own sector. There is no cheaper product-qualified traffic in the acquisition mix, and most companies running attribution have never measured it, because they filed the whole thing under SEO.

Category definition. At scale, “Powered by X” stops describing a vendor and starts naming a category — the brand becomes the default noun for the thing it does, visible as branded search volume and as the phrase competitors must define themselves against. That asset accrues to whoever appears on the most surfaces, which is what a free tier with required attribution buys.

Entity reinforcement. Consistent co-occurrence of your brand with a category description across thousands of independent domains corroborates how search and AI systems model what your company is. As covered in our guide to what backlinks are and what they actually signal, the mention and the association do work that is separable from any equity the link does or does not pass.

What it does not buy. PageRank. Design the system as though the link passes zero, because in every configuration you should be running, it does.

Measure it as a loop, not as a link channel

The reframe that makes attribution manageable: it is a growth loop, not a link tactic. Impressions produce clicks, clicks produce signups, signups produce installs shipping more attribution — a compounding mechanic with a measurable coefficient, belonging next to your other acquisition channels rather than in an SEO dashboard.

Start measuring	Stop measuring	Why
Attribution-sourced signups and conversion rate	Attribution referring domains	The first is the channel working; the second is a churn report misread as growth
Referral sessions from attribution UTMs, by tier	Followed-vs-nofollow ratio	They should all be nofollow. There is no ratio to optimise
Branded search volume and unlinked mentions	Anchor distribution across installs	Uniform branded anchors are correct here; variation is the red flag
Loop coefficient: installs per hundred impressions	Install count as a link metric	One is a growth rate; the other inversely tracks quality

For wider context, our link building statistics for 2026 covers what the acquisition benchmarks look like across tactics, and the review of the best link building tools explains why most backlink platforms report attribution links they should discount — they see the link, not the plan behind it.

Implementation: the attribution component

All of it resolves into a little markup. Here is the Cell A version, still shipping in a great many products.

<!– Cell A: hardcoded, followed, keyword-rich, no owner control –>

best online booking software UK

</a>

</div>

Three violations in six lines: a keyword-rich anchor, a followed link in every install’s template, no mechanism for the owner to qualify it. Now Cell D.

<!– Cell D: branded anchor, nofollow by default, visible,

removable, points to a real product page –>

<a href=”https://example.com/?utm_source=attribution

&utm_medium=embed&utm_campaign=powered_by”

rel=”nofollow”

data-bw-role=”attribution”>Example</a>

</div>

The rel attribute is the compliance fix. The UTMs are the measurement fix, and they matter more than they look: without them your highest-intent traffic arrives as direct or referral noise and is never credited. The version marker is the reproducibility fix.

Expose the control in settings rather than burying it. One toggle, on every plan including free, letting a customer qualify or remove the link is what moves you decisively outside the terms-of-service clause.

Server-rendered or client-injected?

Attribution injected by JavaScript may be crawled inconsistently. That changes nothing about compliance — an uncrawled link is not safer, just invisible — but your backlink tools will report a fraction of the true install base. It is the usual explanation when a product team’s install count and an SEO team’s referring-domain count differ threefold.

A trap worth naming

Having discovered that gap, the tempting next move is to switch attribution to server-side rendering so the links become countable.

Resist it. You would be re-engineering the product to improve a metric you have just finished agreeing does not matter. If there is a genuine product reason to server-render, do it for that reason. Crawlability of a nofollow link is not one.

That general case is worth reading properly in our guide to technical SEO and link building, which sets out where rendering decisions genuinely do change link outcomes — and this is not one of those places.

Cost, failure modes, and the point at which you stop

Attribution reads as free because no media budget touches it. Pricing it honestly stops it becoming a permanent unexamined line item.

Build cost at volume. The markup, a settings toggle across plan tiers, UTM handling and event tracking to signup runs roughly 3–6 developer days — at UK contract rates near £400–600 a day, £1,200–£3,600 initially, plus half a day per quarter for regression checks. Budget the maintenance: attribution silently breaks in redesigns.
Failure mode one: phantom inventory. Customers hide attribution with CSS overrides, and ad blockers strip elements matching common credit-block patterns. Your reported count becomes fiction in a direction you cannot see.
Failure mode two: the legacy template. The new component ships nofollow while an older embed path, email footer or white-label build still carries the original hardcoded followed link. You believe you are in Cell D while a meaningful share of your install base sits in Cell A. Audit every surface, not just the current one.
Failure mode three: UTM collision. Attribution parameters overlap with a paid campaign’s and the channels contaminate each other, usually flattering paid. Namespace them.
Failure mode four: iframe burial. Attribution inside a third-party iframe is invisible to crawlers and to your analytics. Users still see it, so the acquisition value holds — but stop counting it as anything else.
Reproducibility. Version the component and log which release shipped which rel value and anchor, alongside a dated register of the plan matrix. Eighteen months on, when a cohort looks strange, that register is the difference between a two-hour explanation and a two-week investigation.

Failure threshold, and the cheaper fallback

Threshold: if after 90 days attribution has produced no assisted signups and no movement in branded search, the loop is not working.

The wrong response is a followed link or a richer anchor. That does not fix a broken loop; it converts a dead channel into a live liability.

Re-scope or retire. The cheaper fallback is well established: put the same engineering budget into a hosted tool or dataset you own and promote, earning editorial links you selected, from domains you chose, lasting years rather than months.

When not to do this at all

Attribution suits a narrow class of product, and the honest version of this guide says so.

No customer-facing surface. Internal tooling, back-office systems and most APIs have nowhere to put attribution a third party will see. Do not invent a surface: a credit line nobody views is the compliance question with none of the return.
Enterprise-weighted revenue. If your book is concentrated in large accounts, white-labelling is table stakes and requiring attribution costs deals worth more than any link value. The pricing decision is correct even at zero attribution.
Regulated sectors. UK financial services, healthcare and legal clients often have their own constraints on third-party branding and outbound footer links. Ask before you ship; a post-launch compliance objection costs more than the feature.
A profile already dominated by templates. If sitewide boilerplate already accounts for most of your referring domains, more of it does nothing for authority and makes the profile harder to read. Spend the effort on link building strategies that produce editorial links instead.

Opportunity cost is the real argument in each case: attribution competes for engineering time against product work, and should lose unless the loop genuinely closes.

Worked example: a UK booking platform

Anonymised and lightly rounded; the structure is real.

A Bristol booking platform serving salons and independent clinics ran about 2,400 active accounts, roughly 1,780 on a free plan. That plan required attribution: a credit line hardcoded into the booking widget footer reading “Powered by [Brand] — online booking software”, followed, with no setting anywhere to change it.

Their backlink tool reported 1,120 attribution referring domains — well under the install count, because the widget rendered client-side on many sites. The SEO lead had reported that figure as the programme’s largest link source for two years.

Diagnosis

Cell A on every axis: required by the plan terms, hardcoded, followed, keyword-rich anchor. Three spam-policy clauses engaged at once — contractually required without the choice to qualify, keyword-rich links distributed through a widget, and widely distributed template links.

The retention data exposed the second half. Free-tier churn measured 7.4% a month, putting the half-life of those links near nine months; two-thirds of any cohort was gone within a year. The number reported as a growing asset was a treadmill — new installs replacing dead ones at roughly the same rate, the total masking both.

What changed

The next release shipped rel=”nofollow” on the attribution link and changed the anchor from the keyword string to the brand name alone.
A settings toggle was added on every plan including free, letting any customer remove the link or leave it. This was the decisive change: it moved the implementation out of the terms-of-service clause, because the choice now existed.
Attribution stayed mandatory-by-default on the free tier. Nothing in the policy required giving that up, and default-on distribution was commercially valuable.
UTMs were wired through to signup, and attribution moved out of the SEO dashboard into the acquisition report.

Outcome

Reported attribution referring domains fell, partly through opt-outs and partly because the team stopped counting surfaces that had never been meaningful. Expected, and the correct direction.

The finding that mattered came from the new tracking: attribution-sourced signups were about 11% of free-tier acquisition and had never been credited to the channel, landing in direct traffic for years. Branded search grew modestly over two quarters. Rankings for the core commercial term did not move.

The line to take away

Removing the followed links cost nothing in rankings, because they had never passed anything.

It revealed an acquisition channel worth roughly a ninth of free-tier signups that nobody had measured — while retiring a compliance exposure and a vanity metric in one release.

The change did not cost them links. It revealed that they had never had them.

A 90-day plan

If your software renders on a customer’s domain, this is the order.

Days 1–30: establish where you actually are

Inventory every surface that renders attribution: the main product, legacy embed paths, email footers, exported reports, white-label builds. Most teams find one they had forgotten.
For each, record the rel attribute, the anchor, and whether a customer-facing control exists. Place it on the Control Map.
Pull free-tier and paid-tier monthly logo churn and compute the half-life for each cohort. A fifteen-minute calculation that reframes the channel.
Compare reported attribution referring domains against the true install count. The gap is your rendering and blocking loss. Our guide to what link building is and how link value is assessed is the reference point for what the surviving links are worth.

Days 31–60: ship the fix

Change the anchor to a plain brand name first — cheapest, needs no product decision, closes the most-cited failure mode.
Ship nofollow across every surface in one release. A partial rollout leaves you unable to say which cell you are in.
Add the customer-facing control and document it in the help centre. The documentation is part of the fix — a choice nobody can find is weak evidence that one exists.
Wire UTMs and attribution clicks through to signup events before anything else, so you have a baseline.

Days 61–90: re-measure and decide

Replace attribution referring domains in every report with attribution-sourced signups and referral sessions by tier.
Route the attribution-domain trend to whoever owns retention, labelled as what it is: a lagging churn indicator.
Apply the failure threshold. No assisted signups and no branded-search movement at 90 days means the loop is not closing — re-scope or retire, and do not respond by making the link followed.
Add a standing quarterly attribution check to release QA, and keep the versioned register of what shipped when.

In that order the exercise takes one engineer a handful of days and settles a question most software companies have carried unexamined for years: whether the largest block of links in their profile was ever a link asset. The honest answer is usually no — and the acquisition channel underneath it is a better outcome than the links would have been.

Open Source as a Link Strategy: Earning Links by Shipping Free Software