Algorithm Impact Example: -23% Traffic, Core Update Confirmed
This example shows the aiwebpageseo Algorithm Impact Monitor for a SaaS marketing site after a 28-day -23% organic traffic drop. Confirmed correlation with Google Core Update rolling 2026-05-08 to 2026-05-18. Category volatility peaked at 8.4. Informational blog content hit hardest (-42%); transactional pages held steady. AI Overview coverage rising on 38% of tracked queries.
| Date | Event | Traffic delta | Category volatility |
|---|---|---|---|
| 2026-04-15 | Baseline normal | — | 2.1 (low) |
| 2026-05-08 | Google Core Update started | -4% | 5.2 (elevated) |
| 2026-05-12 | Update rollout continues | -14% | 8.4 (high) |
| 2026-05-18 | Update rollout complete | -23% | 6.1 (settling) |
| 2026-05-20 | Current | -23% | 3.4 (returning normal) |
Informational blog content — heaviest drop
Blog category lost 42% of organic clicks. Note the limits of that observation: Google folded the helpful content system into its core ranking systems, so there is no separate helpful-content signal whose weight can be inspected, and E-E-A-T is criteria used by Google's human quality raters — not a score computed for your pages. What is defensible is that the drop is concentrated in informational content while transactional pages held, which points at content quality rather than a site-wide technical fault. Action: audit the blog for thin, duplicative and unsourced pages; improve or remove the weakest, and do not expect a per-page recovery you can attribute to any single edit.
Product comparison pages — moderate drop
"X vs Y" pages down 18%. There is no E-E-A-T weighting to point at — no ranking signal of that name exists, and no schema property confers it. What Google's guidance for reviews does describe is evidence of first-hand use: measurements, photographs of the product in use, and a clear account of what was tested and how. Action: show the testing rather than asserting the expertise. Note also that schema.org has no ComparisonTable type — do not invent one; mark up what genuinely exists on the page.
Transactional pages — stable
Pricing, signup, contact pages: traffic stable. Confirms drop is content-quality-driven, not site-wide technical issue.
AI Overview coverage rising on your queries
38% of your tracked queries now show AI Overview. Where AI Overview cites you (12% of those): traffic neutral. Where it cites competitors (88%): traffic -45%. Long-term: pivot informational content to commercial intent.
| Action | Effort | Expected lift | Recovery window |
|---|---|---|---|
| Audit blog & remove thin content | Medium | +8 to +12% | 6-9 months |
| Add named experts + credentials | Medium | +4 to +8% | 4-8 months |
| Pivot comparison pages with first-hand evidence | High | +6 to +10% | 4-8 months |
| Diversify channels (AI engines, social, email) | High | +15 to +25% | 3-6 months |
| Wait for next core update | Low | +/-5% | 3-6 months |
The timeline is about as clean as this evidence ever gets. Traffic is flat on 15 April. It turns down on 8 May, the day the rollout begins. It deepens through the rollout, bottoms at −23% on the day the rollout completes, and category volatility rises and falls on the same curve. Transactional pages hold steady throughout while informational pages fall hard.
That pattern is strong evidence that the core update caused the loss. It is not evidence of why. Google publishes no per-component breakdown of a core update and tells no one which of its systems affected which pages, so every statement about mechanism — including the ones in the report above — is inference from the shape of the damage. This is worth holding onto, because the industry's instinct after every update is to name a cause with more confidence than the data supports, and then to spend six months acting on it.
There is also a second thing happening in the same window. AI Overviews now appear on 38% of your tracked queries, and where they cite competitors your clicks fall sharply. That is an independent change in the results page arriving at the same time as the update. If you attribute the whole 23% to the update, you will fix the wrong thing; if you attribute it all to AI Overviews, you will do the same. Both are true and they need separating before you plan anything.
A core update is a reassessment, not a sanction
Nothing was applied to your site. A penalty — a manual action — is a specific thing that appears in Search Console with a named reason and an appeals process. You have not received one. What happened is that Google changed how it weighs the pages competing for your queries, and your pages came out lower relative to the others. That distinction is not pedantry: a penalty is something you get lifted, a reassessment is something you have to win back, and the two demand entirely different work.
There is no E-E-A-T value attached to your pages
E-E-A-T is criteria used by Google's human quality raters when they assess whether search results are good. Their judgements are used in aggregate to evaluate ranking changes; they are not computed for your URLs and no property of your markup confers them. Any tool reporting your E-E-A-T as a number has invented that number. What the underlying guidance describes — demonstrable first-hand experience, real subject knowledge, an identifiable and accountable author — is worth building. Just do not imagine you are moving a dial, because there is no dial.
The same applies to the helpful content system: it was folded into the core ranking systems, so there is no separate signal to have been weighted up in this update. When a report says "helpful-content signals were weighted higher", it is describing a mechanism nobody can observe. The observable fact is narrower and still useful — informational pages fell, transactional pages did not.
The most expensive mistake after an update is rewriting 200 pages on a hypothesis nobody tested. Four checks, in this order, cost a day and rule out most of the wrong explanations.
Confirm it is not technical
Check that the affected pages still return 200, are still indexed, still carry a self-referential canonical, and were not accidentally noindexed by a template change in the same window. A deployment that coincides with an update is the single most common false attribution in this whole field, and it is the cheapest thing to eliminate.
Separate impressions from clicks
A page that lost rankings shows falling impressions. A page that kept its position but lost clicks to an AI Overview shows steady impressions and a collapsed click-through rate. These are different problems and Search Console distinguishes them cleanly. Given that AI Overviews now appear on 38% of your queries, expect to find a mix — and expect the mix to change what you do.
Look at query intent, not page type
The blog dropped 42% and pricing held, which suggests the reassessment hit informational intent specifically. Check whether the queries you lost were ones where a searcher wanted an explanation rather than a product. If so, the competitors who took those positions are the ones to read.
Read the pages that replaced you
Not their backlink profile, not their schema — their content. In most core-update analyses the honest finding is that somebody wrote a better, more specific, more evidently first-hand answer than yours. That is unglamorous, unautomatable, and correct.
The percentages in the recovery forecast are scenario ranges, not predictions, and they should be presented internally as such. Nobody can forecast recovery from a core update, because recovery depends on how much everyone else improves in the same period and on when the next update runs — neither of which is knowable.
Google's own position is that sites hit by a core update may see no change at all until a subsequent core update runs, which can be months away, and that improvement is not guaranteed even then if competing pages have got better faster. Any agency proposal containing a recovery timeline expressed to the month is selling certainty it does not have.
What the table is actually good for is ordering by effort and by confidence. Removing genuinely thin content is high-confidence, low-effort work that is correct regardless of the update. Rebuilding comparison pages around evidence you actually gathered is high-effort and high-confidence. Waiting is free and does nothing. Channel diversification is high-effort, addresses a real concentration risk, and does not repair a single page that dropped — which is why it belongs in the strategy conversation and not in the recovery plan.
Thirty-eight percent of tracked queries now show an AI Overview. Where you are cited, clicks are roughly neutral. Where a competitor is cited, clicks fall sharply. This is a change in the results page, not a change in how your pages are ranked, and it demands a different response.
Two corrections are worth stating, because the surrounding advice is usually wrong. There is no opt-in, no schema type and no file that makes you eligible to be cited in an AI Overview — eligibility follows from ordinary Googlebot access and ordinary Search eligibility. And Google-Extended has nothing to do with it: that directive governs whether your content trains Gemini, and disallowing it costs you nothing in Search or in AI Overviews. Sites that blocked Google-Extended in a panic after an update lost precisely nothing, and gained precisely nothing either.
The realistic response is uncomfortable but simple. For queries where an assistant can fully answer the question in three sentences, the click was probably never very valuable and is not coming back. For queries where the answer is genuinely involved — where the reader needs the caveats, the numbers or the method — the page that supplies the best version of that answer is the one that gets cited and the one that still earns the click. Write for the second category and stop competing for the first.
How confident can I be that a core update caused this drop?
Reasonably, but not completely, and the distinction matters. The evidence here is circumstantial and it is the strongest kind available: the decline begins on the day the rollout began, deepens through it, and stabilises when it completes; category volatility rises and falls on the same curve; and the loss is concentrated in one content type while transactional pages hold steady. That pattern is very hard to produce by coincidence. What it does not give you is a mechanism. Google does not tell anyone which components of a core update affected which pages, so any statement about why is inference. Treat the correlation as established and the explanation as a hypothesis.
Was this the helpful content update penalising my blog?
There is no separate helpful content system left to penalise you. Google folded it into the core ranking systems, which means there is no distinct signal whose weight can be examined and no switch that was flipped. It also means there is no penalty in the technical sense: a core update is a reassessment of how pages are ranked relative to each other, not a punishment applied to a site. Your blog did not receive a sanction. It was re-evaluated against everything else competing for those queries and came out lower, which is a different problem with a different solution.
Can I improve my E-E-A-T to recover?
Not directly, because E-E-A-T is not a thing your site has a value for. It is a framework Google's human quality raters use when judging whether search results are good, and rater judgements are used to evaluate ranking changes in aggregate — they are not applied to your URLs as a score. No schema property confers it, adding author bios does not compute it, and any tool reporting your E-E-A-T as a number has invented that number. What the underlying guidance actually describes is content that shows first-hand experience, demonstrates real knowledge, and is produced by someone identifiable and accountable. Build that, because it is what good content looks like — not because a score will move.
Should I delete the pages that lost the most traffic?
Only if they are genuinely bad, and losing traffic is not on its own evidence of that. Deletion is a blunt instrument that permanently discards any links, history and residual value a page holds, and a page that dropped in a reassessment may recover in a later one. The honest test is to read the page and ask whether it is the best available answer to the question it targets. If it is thin, duplicative or says nothing a dozen other pages do not say better, improving it is the first option and removing it is the second. If it is good and it dropped anyway, deleting it fixes nothing and destroys something.
How long does recovery from a core update take?
Nobody can tell you, and any timeline expressed with confidence should be treated as marketing. Google's own guidance is that sites affected by a core update may not see change until a subsequent core update runs, which can be months away, and that there is no guarantee of recovery even then if competing pages have simply improved more. The recovery estimates in the table above are scenario ranges, not forecasts: they express how much of your loss each action could plausibly address, and they are all conditional on the work being genuinely good. Plan on the assumption that improvement is slow, non-linear, and impossible to attribute to a single change.
What does the AI Overview data actually tell me?
That a growing share of your queries now display an AI Overview, and that when it cites competitors instead of you the click loss on those queries is severe. That is a real and measurable effect on your traffic, and it is separate from the core update even though the two arrived together — which is precisely why the two must not be conflated when you attribute the 23%. Being cited in an AI Overview is a function of ordinary Googlebot access and ordinary Search eligibility; there is no separate opt-in, and Google-Extended has nothing to do with it, since that controls Gemini training only. The practical response is the same as for the update: be the page that answers the question best, in prose that can be lifted and still make sense.
Is diversifying away from Google the highest-value action on the list?
It is the one with the largest range on the table, and that range should be read as uncertainty rather than promise. Reducing dependence on a single channel is sound strategy for reasons that have nothing to do with this update, and if a single algorithm change can remove a quarter of your traffic, your risk was already concentrated. But the effort is high, the payback is slow, and it does not repair the pages that dropped. Do it because concentration risk is real, not as a way of avoiding the content work — the content work is what recovers the traffic you lost.
Related Demo Reports
Run Algorithm Impact Monitor on Your Own Site
Get your real algorithm impact diagnosis with timeline correlation, page-type breakdown and recovery forecast by remediation action.