Backlink history: tracking links over time and relating them to rankings and AI citations

You manage link building for a B2B company and have to explain why a page climbs, drops or holds steady. Backlink history tracks your links over time: it reads your link profile as a time series of links gained, lost or changed, each one dated, then relates them to rankings and AI citations. It provides the proof, the last stage of the process described in our link building guide.

You will find what to track and how often, how to qualify a loss, and the protocol that connects a link to a change without confusing correlation with causation.

Why read your backlinks over time

A snapshot of your profile at a single point in time shows what you have, not what changed. A ranking drop is explained by a dated event: a wave of losses, an attribute that switches, a source page that returns a 404. Without history, you only see the drop after it has happened, with nothing to tie it to.

Reading over time tracks gains, losses and the net change. It protects the links that were hard to earn, spots a breakpoint before it turns into a prolonged decline and proves what actually changed.

What is backlink history?

It is the chronological tracking of your profile, on two levels: an aggregated view (unique dofollow referring domains, anchors and target pages, period by period) and a link-by-link view (status, dates, anchor text, URL, attribute).

A healthy profile is recognised by explainable stability:

• Steady growth in dofollow referring domains, or step changes that match your actions.
• Controlled losses, never clustered without an identifiable cause.
• A stable attribute mix, without sudden swings.
• A stable anchor profile across brand, partially optimised, exact-match and generic anchor text.
• Target pages spread according to your priorities.

A lost link is not a toxic link

A link that disappears lowers the net change without putting the site at risk. A toxic link stays in place and is a problem because of what it is (questionable domains, repetitive anchors). History helps you date these signals, but dealing with them belongs to backlink analysis, which keeps disavow for manual actions and for link schemes you created yourself.

Reading a competitor domain's history

A link index also keeps the links detected in the past on competitor domains: you can see how fast they gain referring domains and on which pages. The useful question becomes: are we earning dofollow referring domains faster than the pages ranking above us?

Setting up the time series: what to track and how often

A usable time series tracks few objects, always the same ones, at a fixed cadence: weekly during a campaign, monthly when acquisitions are low and stable. The table connects each tracked object to the decision it triggers:

Record dated marketing events (publication, campaign, redesign) in the same series: without those markers, a link curve explains nothing.

The fields to keep for each link

To rebuild a reliable timeline, archive stable fields, link by link:

• source URL and target URL;
• anchor text and anchor type;
• attribute: dofollow, nofollow, sponsored or ugc;
• first-seen and last-seen dates;
• status: live, lost or not verified;
• context: campaign, publisher, link owner.

Gross, net and recoveries

In the timeline, separate four flows to avoid false alarms:

• Gross gains: links detected for the first time during the period.
• Gross losses: links not found at the latest check.
• Recoveries: links lost then found again, to be taken out of the losses.
• Net change: gains minus losses, calculated only after each loss has been qualified.

30-, 90- and 180-day cohorts: measuring link lifespan

Durability is the most underestimated indicator. Group links by cohort, that is by month of acquisition, and measure the share still present after 30, 90 and then 180 days. A cohort that shrinks fast points to unstable publishers; a stable cohort shows the formats and sites worth using again.

A highly volatile profile creates "yo-yo" performance, where the effect of an acquisition is confused with the effect of a loss. Compare cohorts by publisher type, campaign and target page rather than judging a single link.

Velocity and diversification

Over the same periods, monitor these distributions, which tell you whether growth is healthy:

• Acquisition pace: a spike without a dated event should trigger a check of sources and anchors.
• New domains or repeats: the target is counted in unique dofollow referring domains, not in links.
• Concentration on one URL: a loss on the page that receives everything is costly.
• Followed vs nofollow mix: a rapid shift signals a publisher policy change or a link reclassified as sponsored.

Qualifying a loss before acting

A loss is only actionable if you know what disappeared, when, and which target page the link supported. Not every loss is a removal:

Before you conclude a loss is real, check that the source page is still indexed and accessible, and that the link actually appears when the page is rendered: otherwise, you are dealing with maintenance, not strategy.

Assessing criticality, then deciding

A loss becomes a risk of a sustained drop when it affects a strategic page, a domain that is rare in your profile and a contextual link, and impressions fall over the same window. It remains low-impact on a repeat link whose page stays stable. The action follows this diagnosis:

• Recover: fix a 404, request reinstatement of a link removed by mistake.
• Compensate: look for a new referring domain if the source page has disappeared.
• Rebalance: diversify target pages and anchors if the loss reveals a concentration.
• Prevent: document volatile publishers and set alerts for critical links.

Lost links and the guarantee

A link bought through a platform that offers a 24-month guarantee is tracked with its status: presence verified, guarantee active or expired. While the guarantee runs, a loss is handled within that framework first, before any new purchase. A link imported from another provider is tracked in the same place, but outside the guarantee.

Connecting history to rankings: the correlation protocol

A link is judged by the page it strengthens, not by the domain curve. The protocol compares each page with itself, before its first link and today, cross-referencing rankings, impressions and AI citations. The result is a time-based correlation, never causation: it tells you what moved at the same time, not what caused it.

Traffic is not enough as a measure: 60% of searches end without a click (Semrush, 2025). Track rankings and impressions in Google Search Console, which also shows external links, anchor text and referring sites, and track conversions in Google Analytics 4.

The steps of the protocol

The protocol is applied page by page, with a fixed 28-day window:

• Choose the pages: the URLs strengthened by a campaign, pillar pages and business pages first.
• Freeze the baseline: rankings, impressions, clicks and citations recorded over the 28 days before the first link.
• Log events: gains, losses, attribute changes, but also redesigns and seasonality.
• Compare equal windows: the last 28 days against the baseline, for the same page and the same keywords.
• Look for patterns that repeat across several strengthened pages.

Reason in batches: a single link gets lost in SERP noise, whereas a batch of similar links (same publisher type, same period) isolates an effect better.

Reading the common scenarios

The comparisons reveal recurring situations:

• Growth without link gains: improved content, better intent match or rising demand.
• Decline after a loss: common on a business page when the lost link was dofollow and contextual.
• Time lag: a link change takes time to be crawled and reassessed.
• Cumulative effect: several small losses over a few weeks weigh more than a single event.

Conversely, a loss with no effect affects a link that was weak, not indexed or offset by other gains.

Measuring AI citations before and after the first link

ChatGPT, Perplexity, Gemini and AI Overviews cite sources and name brands: mentions and links earned on domains they cite also count. To check this, apply the same discipline as in SEO:

• set a list of questions, for example 30 to 50, and keep the prompts unchanged;
• record the baseline before the first link: is your brand cited, and which pages are;
• record again at a fixed cadence;
• avoid adding a redesign, new links and new content at the same time.

Each question then gets a status, read page by page:

A gained question remains a correlation: nothing allows you to state that a link produced the citation. Naming is not linking: the brand mention carries the GEO signal, the link carries the SEO signal, and the two add up on a domain cited by AI engines.

Archiving and alerting

Without archiving, a simple question goes unanswered: what changed during the week of the decline? Archiving turns intuition into a verifiable diagnosis, with a few rules:

• Timestamping of every check, with first and last detection.
• Change log: anchor, attribute, target URL and status, before and after.
• Evidence: periodic exports and control notes, useful in case of disputes.
• Action record: date, strengthened pages, publisher type, hypothesis, result at day 28 and day 56.
• Retention: at least 12 to 24 months to capture seasonality and cycles, and the full length of any guarantee.

The alerts to set

Keep alerts for the events that change a decision:

• loss of a link to a page you want to strengthen or defend;
• loss of a referring domain that is rare in your profile;
• a wave of losses within the same week;
• an attribute or anchor change on a purchased link;
• a source page that switches to 404, redirect or noindex.

Depending on your situation

Depending on what your history has taught you, the next step is here:

• You need to choose the pages to strengthen: the backlink strategy sets out the five diagnostic states and the link target per URL.
• You are preparing new links: the link building campaign defines the publishers, anchors, brief and a spread-out pace.
• You need to replace lost links: the eligibility grid for buying backlinks includes the guarantee and the price per Trust Flow point.
• You track your presence in AI answers: see how links and mentions count for generative engines.

To keep this history without a spreadsheet, the reporting in the Incremys Offsite media module tracks, at the time of writing, the presence of each link (live, lost, not verified), the guarantee status of links bought through the platform, Trust Flow, the anchor profile, rankings and ChatGPT citations gained or lost per question. Each page is compared with itself before its first link and today.

FAQ on backlink history

Short answers to the most common questions about tracking links.

What is backlink history?

Backlink history is the chronological tracking of a link profile: links gained, lost or changed, with their dates, anchors, attributes and statuses. It is used to explain a change in visibility by a dated event and to relate links to rankings and AI citations.

Why do you lose backlinks?

You lose backlinks when the source page is deleted, modified or de-indexed, when the link changes attribute or URL, or when your target page returns a 404. Before concluding it is a removal, rule out technical causes: redirects, canonicals, link rendering and source-page accessibility.

How do you connect a backlink to a ranking increase?

You connect a backlink to a ranking increase by comparing the page with itself, before its first link and today, over two 28-day windows. Log the other changes and reason in batches. The result remains a time-based correlation: it never proves that the link caused the increase.

Can you measure the effect of a link on ChatGPT?

You can measure how ChatGPT citations evolve, not the specific effect of a link. Set a list of questions, record them before the first link and then at a fixed cadence, and classify each question as gained, lost, stable or no baseline. A question not tracked before the first link has no "before", and a gained citation remains a correlation.

Should you replace a lost link?

Not systematically: qualify the loss and its criticality first. A 404 is fixed on your side; an accidental removal is reported to the publisher. If the link supported a page you want to strengthen and the source page has disappeared, look for a new referring domain. A link bought under guarantee is handled within that framework first.

How long should you keep backlink history?

Keep at least 12 to 24 months of history. That depth captures seasonality, lets you read 180-day cohorts and covers the length of a 24-month guarantee. Keep exports in the same format so you can compare periods.

What should you do when a link changes anchor text or attribute?

Record the change with its date, the before and after state and the source URL, then assess its criticality. A link that has become nofollow leaves the dofollow referring domain count, but can still count for GEO if the domain is cited by AI engines. On a purchased link, check that the change respects the agreement made with the publisher.