The short answer
Should a software company invest in a link roadmap or buy single links?
Invest in a roadmap when durability is the question. Ahrefs (2022) found that 66.5% of links pointing to a sample of 2,062,173 websites had rotted since January 2013, so one link is an asset that decays. A roadmap plans for that decay: it concentrates links on pages that earn revenue, reserves budget to replace losses and checks every placement on a schedule. A single purchase makes sense only as a labelled sponsorship or to fill one defined gap on one page.
What does each model actually buy?
A single link purchase is one placement acquired as a transaction. Nothing connects it to a target page, a gap analysis or a monitoring plan, so you cannot tell afterwards whether it did anything.
A strategy roadmap is a plan that fixes four things before outreach starts: the target pages, the link gap, the publisher qualification rules and the review schedule. It treats every placement as something that has to be when it fails, replaced. For the budget side of this decision, see Scaling Link Building Budgets. This article covers the other half: how long the result lasts.
Why does a single link decay?
Ahrefs’ link rot study (2022) measured links pointing to 2,062,173 sampled websites from January 2013 onwards. It reports that 66.5% rotted, that another 6.45% showed temporary errors and that a further 1.55% had issues which stopped them counting for ranking, which brings the total counted as lost to 74.5%. Claim strength: measured by Ahrefs, 2022, updated 2024.
The reasons matter more than the headline. Of the links counted as lost:
| Why the link was lost | Share of lost links | What it means for a purchased placement |
|---|---|---|
| Page dropped from Ahrefs’ index (not crawlable, not indexable, or the domain is gone) | 47.7% | The placement may sit on a page that no longer ranks or exists. This is the case an indexation check at 90 days catches |
| Link removed, page still live | 34.2% | An editor changed the article. Content refreshes and policy changes cause this |
| Crawl error | 6.45% | Unclear. Ahrefs does not count these in its rot figure |
| Redirect (301/302) | 5.99% | The page moved. The link may still exist elsewhere |
| Not found | 4.11% | The linking page was deleted |
| Not canonical | 0.82% | Usually a URL standardisation, not a loss |
| Noindex | 0.73% | The link exists but the page is hidden from search |
Source: Ahrefs (2022) link rot study. Shares are of lost links, not of all links.
What the study cannot tell you. It samples links to random websites over nine years. It does not isolate editorial placements bought by software companies, so it cannot give you a durability rate for the links you buy. Treat 66.5% as evidence that decay is the default, not as a forecast for your portfolio (Ahrefs, 2022).
What we see in our own tracking. Placements we track tend to stay live beyond 365 days. That is an observation from our internal data, it is not common across the market, and it is not a benchmark you can apply to other vendors. We publish no figures for it. Claim strength: observed, qualitative only.
What does a roadmap change in practice?
Three things change, and each answers a failure mode in the table above.
- Qualification before purchase. Pages that are dropped from the index are the largest loss category, so qualify publishers on indexation and real organic traffic before any link is placed. The site qualification checklist sets this out step by step.
- Checks at 90, 180 and 365 days. Each check confirms three facts: the page is live, it is indexed, and it still links to the right URL. A link that fails is reclaimed or replaced, which is cheaper than finding out a year later.
- A replacement reserve. Budget for some placements to fail. The right size is a planning assumption you set from your own tracking, not a published rate. Claim strength: opinion.
What do AI answer engines add to the case?
They weaken the case for isolated purchases. Ahrefs’ December 2025 study of 75,000 brands (Linehan, 2025b) compared brand-level signals with how often ChatGPT, Google AI Mode and Google AI Overviews mention a brand. YouTube mentions correlated most strongly (about 0.737), and branded web mentions followed (0.656 to 0.709 depending on the platform).
Domain rating was a mid-tier signal (0.266 to 0.326), and Ahrefs describes the link-metric correlations as very weak and the site page count as almost unrelated (about 0.194). In Ahrefs’ earlier study of AI Overviews alone, branded web mentions correlated 0.664 with a brand appearing, against 0.218 for the number of backlinks (Ahrefs, May 2025 (Linehan, 2025a)). Claim strength: derived from correlational studies. Ahrefs states that correlation is not causation.
Two limits apply. The study measures brand mentions, not links you bought. And it does not say links are useless: classic rankings still respond to them, and Ahrefs’ own advice (Linehan, 2025b) is that low-quality links will not help. The practical reading is narrow: if the goal is AI visibility, an isolated bought link is the weakest tool in the set, and being named across many credible sites is a stronger target. Our earlier piece on brand mentions and AI answers covers that side.
What correlates with being named by AI answer engines?
Correlation with brand mentions in AI answers, by platform. Pick a platform. Higher means a stronger relationship. This is correlation across brands, not proof of cause.
Source: Linehan (2025b), Ahrefs Blog, 12 December 2025, 75,000 brands, Spearman correlation. YouTube mentions and site pages are Ahrefs' headline values (about 0.737 and about 0.194), shown for all three platforms. Backlink metrics are described by Ahrefs as very weak and are not plotted because no figure was verified.
Text version of the chart (correlation with AI brand visibility, Ahrefs, December 2025, 75,000 brands): (Linehan, 2025b)
| Signal | ChatGPT | AI Mode | AI Overviews |
|---|---|---|---|
| YouTube mentions | about 0.737 | about 0.737 | about 0.737 |
| Branded web mentions | 0.656 | 0.709 | 0.664 |
| Branded anchors | 0.511 | 0.628 | 0.527 |
| Branded search volume | 0.352 | 0.466 | 0.392 |
| Domain rating | 0.266 | 0.285 | 0.326 |
| Number of site pages | about 0.194 | about 0.194 | about 0.194 |
YouTube and site-page figures are the headline values Ahrefs reports across the three platforms. Ahrefs’ chart of YouTube mentions and mention impressions spans 0.712 to 0.740 across ChatGPT, AI Mode and AI Overviews, so the “about 0.737” is a rounded headline, not a per-platform value (Linehan, 2025b).
Are purchased links allowed?
Google’s spam policies (2026b) list buying or selling links for ranking purposes as link spam. The same page says buying and selling links is a normal part of advertising and sponsorship, and that it is not a violation when such links carry a rel=“nofollow” or rel=“sponsored” attribute.
That splits “single link purchase” into two different things. A labelled sponsorship on a publication your buyers read is a media buy: judge it on audience, not on ranking effect. A placement bought to pass ranking credit without a label is the pattern the policy targets. We do not recommend the second. Our explainer on buying links goes through the distinction.
When is a single purchase the right call?
Only in three situations:
- You want exposure with a specific audience and the link is labelled as sponsored.
- One target page has a defined, small gap and you have already qualified the publisher.
- You are testing a publisher before committing a larger budget. Keep the test to a small batch and measure it at 90 days.
Reclaiming an unlinked brand mention is not a purchase, and it is usually the cheapest link available.
How do you choose?
| Question | If yes | If no |
|---|---|---|
| Do you know which pages the link should support? | Roadmap | Fix this first. Without a target page, any link is noise |
| Can you check the placement at 90, 180 and 365 days? | Roadmap or a tracked single buy | Do not buy. You cannot learn from it |
| Is the link labelled as sponsored? | Treat it as media spend | Review the policy risk before going ahead |
| Is the main goal AI visibility? | Prioritise mentions across credible sites | Links remain relevant for rankings |
Key takeaways
- A link is closer to a lease than a purchase: in Ahrefs' sample, at least 66.5% of links to the sampled sites were gone within about nine years, and 74.5% counted as lost for SEO purposes (Ahrefs, 2022).
- Of the links Ahrefs counted as lost, 47.7% sat on pages that were dropped and 34.2% sat on pages that stayed live but no longer linked.
- That sample covers links to random sites, not purchased placements. Nobody publishes a durability benchmark for bought links, so measure your own at 90, 180 and 365 days.
- For AI answer engines, Ahrefs found brand mentions correlate with visibility far more strongly than domain rating, and it called link and content volume weak signals. Correlation is not causation, but it weakens the case for buying isolated links to win AI citations.
- Google treats buying links for ranking purposes as link spam (Google, 2026b). Sponsorship links are allowed when they carry rel="sponsored" or rel="nofollow".
References
- Ahrefs (2022) 'At least 66.5% of links to sites in the last 9 years are dead', Ahrefs Blog, 29 April (updated 2 February 2024). Available at: https://ahrefs.com/blog/link-rot-study/ (Accessed: 20 October 2026).
- Google (2026b) 'Spam policies for Google web search', Google Search Central, last updated 28 August. Available at: https://developers.google.com/search/docs/essentials/spam-policies (Accessed: 20 October 2026).
- Linehan, L. (2025a) 'An analysis of AI Overview brand visibility factors (75K brands studied)', Ahrefs Blog, 26 May (updated 27 April 2026). Available at: https://ahrefs.com/blog/ai-overview-brand-correlation/ (Accessed: 22 October 2026).
- Linehan, L. (2025b) 'Top brand visibility factors in ChatGPT, AI Mode, and AI Overviews (75k brands studied)', Ahrefs Blog, 12 December (updated 12 August 2026). Available at: https://ahrefs.com/blog/ai-brand-visibility-correlations/ (Accessed: 22 October 2026).
