SEO KPIs 2026: 9 metrics that matter as clicks vanish

TL;DR: Sessions and raw click-through rate no longer describe what an SEO program is doing, because 68.01% of U.S. Google searches now end without a click (Search Engine Land). Replace a clicks-only dashboard with nine metrics that separate citation from clicks: share of answer, AI Overview impression share, branded search growth, unlinked brand mentions, assisted conversions, non-branded conversion rate, content decay rate, cannibalization delta, and organic pipeline contribution.

---

Table of contents

1. Why sessions-first reporting stopped working
2. Old KPIs vs 2026 KPIs, compared
3. KPI 1: Share of answer
4. KPI 2: AI Overview and AI Mode impression share
5. KPI 3: Branded search volume growth
6. KPI 4: Unlinked brand mention volume
7. KPI 5: Assisted conversions
8. KPI 6: Non-branded organic conversion rate
9. KPI 7: Content decay rate
10. KPI 8: Cannibalization delta
11. KPI 9: Organic pipeline contribution
12. Building the dashboard: one query set, three surfaces
13. What to delete from your reporting deck
14. Frequently asked questions
15. Key takeaways

Why sessions-first reporting stopped working

Zero-click search hit 68.01% of U.S. Google queries in the first four months of 2026, up from 60.45% in 2024, based on SparkToro’s clickstream analysis reported by Search Engine Land. That is a 7.56-point jump in two years, and it means most searches you rank for never send you a visitor at all.

The drop is not evenly spread. Queries that trigger an AI Overview run zero-click at roughly 83%, and Google’s AI Mode runs closer to 93%, against about 60% for queries without an AI Overview, per the same report. Mobile searches sit at 77% zero-click; desktop is closer to 46.5%. Informational queries run 74% zero-click; transactional queries run 31%. A dashboard that reports “organic sessions” without splitting by query type and device is averaging together traffic sources that behave nothing alike.

Click-through rate has moved with it. Seer Interactive tracked 3,119 informational queries across 42 organizations from June 2024 through Q3 2025 and found organic CTR fell 61% on queries where an AI Overview appears, with paid CTR falling 68% on the same queries (Search Engine Land). Seer’s 2026 update to that study found the drop is not uniform across brands: organizations that get cited inside the AI Overview see 35% higher organic CTR and 91% higher paid CTR on the same queries than organizations that rank but do not get cited (Seer Interactive). Ranking is no longer the variable that predicts your traffic. Citation is.

None of this means SEO stopped working. It means the click stopped being the only unit that matters, and a KPI set built entirely on sessions and CTR will show a program declining even when it is influencing more buying decisions than it did in 2024. The nine metrics below split into two groups: citation-side metrics that tell you whether AI systems and Google surfaces are including you, and business-side metrics that tell you whether the traffic and mentions you do get are turning into revenue. You need both groups, and you need to stop reporting the metrics that only made sense when every search ended in a click. If your traffic already dropped and you need a recovery plan first, start with the 30/60/90 recovery plan keyed to your measured CTR-loss band before building the KPI set below, and if the drop feels more permanent than a dip, the zero-click survival guide covers which business models still work when nobody clicks at all.

Old KPIs vs 2026 KPIs, compared

Metric	Still useful in 2026	Why
Total organic sessions	✗	Masks the 68% of queries that never click; treats informational and transactional traffic as one number
Aggregate keyword rankings	✗	A #1 ranking with no AI citation can still lose 61% of its clicks (Search Engine Land)
Raw impressions in Search Console	✗	Google’s own Generative AI report now shows AI Overview and AI Mode impressions with no clicks, CTR, position, or query data attached (ppc.land)
Share of answer	✓	Measures whether your content is the one AI systems actually surface for your target queries
Branded search volume growth	✓	A demand signal Google reads independently of your rankings
Unlinked brand mentions	✓	Correlates with AI Overview visibility at 0.664 versus 0.218 for backlinks, per Ahrefs’ study of 75,000 brands (Ahrefs)
Assisted conversions	✓	Recovers organic’s contribution to conversions that close on a different channel
Non-branded conversion rate	✓	Isolates whether the traffic you still get converts, independent of volume
Content decay rate	✓	Flags which pages are losing impressions before rankings visibly drop

KPI 1: Share of answer

Share of answer measures how often your brand, page, or product shows up inside AI-generated answers for a defined set of target queries, as a percentage of the total answers you tested. It is the citation-era replacement for share of voice: instead of counting how often you rank in the top ten, you count how often an AI system actually names you when it answers the question.

Build it by running the same 20-50 target prompts against ChatGPT, Perplexity, Google AI Overviews, and Gemini on a fixed cadence (weekly or biweekly), logging whether your domain is cited, whether a competitor is cited instead, and whether no one is cited. Divide citations by total prompts run per engine to get a percentage. Track it per engine, not blended, because citation behavior differs sharply by platform: Reddit was the single most-cited domain across both Google AI Overviews and Perplexity from August 2024 through June 2025, and ChatGPT cites Reddit in roughly 12% of U.S. answers, according to a joint analysis reported by Sitebulb and CMSWire. A page can dominate share of answer on Perplexity and be invisible on ChatGPT for the identical query set, and a blended number hides that.

Share of answer is a leading indicator. It moves weeks before ranking or traffic changes show up, because AI systems re-crawl and re-rank their source sets faster than Google refreshes classic rankings. If you want the mechanics behind why a page ranks on Google but never gets cited by an AI engine, see the diagnostic breakdown of the seven causes and what actually separates SEO from GEO. For the full definition, measurement setup, and reporting cadence, the dedicated guide to share of answer goes deeper than the summary here.

KPI 2: AI Overview and AI Mode impression share

Google launched a Generative AI performance report inside Search Console on June 3, 2026, breaking out impressions from AI Overviews and AI Mode separately from the classic web report, starting with a subset of UK sites under pressure from the CMA before a wider rollout (ppc.land). The report has one hard limitation: it shows impressions only. No clicks, no CTR, no average position, and no query-level breakdown.

That limitation is the point. Impression share, not CTR, is the metric this report supports, because Google is not giving you the data to compute CTR for AI surfaces yet. Track your AI Overview impression trend against your classic web impression trend for the same query set. A widening gap, more AI Overview impressions with flat or falling web impressions, tells you Google is answering more of your queries inside the SERP itself rather than sending the query to the ten blue links at all. That is a structural shift in the query, not a ranking problem, and no amount of on-page optimization fixes it.

Pair this with your own AI Overview presence audit: for your top 50 non-branded queries, record weekly whether an AI Overview appears at all, and if it does, whether you are cited inside it. The trend in “AI Overview present, not cited” is your risk queue. Every query that moves into that bucket is a query where you now compete for the ~17% of clicks that remain after the 83% zero-click rate on AI Overview queries, and you are not even eligible. The event and referrer setup for pulling AI-engine traffic into this same view is covered in the full GA4 tracking guide for ChatGPT and Perplexity referrals.

KPI 3: Branded search volume growth

Branded search volume measures how many people search your product or company name directly, independent of any specific page ranking. It matters because Google reads a rise in branded queries as demand that its algorithms should satisfy in adjacent non-branded queries too, and because branded query volume happens outside the AI-citation fight entirely: nobody needs an AI Overview to tell them what your product is once they already know your name.

Pull it from Search Console by filtering the query report to your brand terms and close variants (misspellings, “[brand] pricing”, “[brand] vs”, “[brand] reviews”), then track the trend as a standalone line, separate from your non-branded query performance. A branded query set that grows while non-branded impressions stay flat usually means offline or upper-funnel activity (a podcast mention, a conference talk, a competitor comparison someone saw) is doing work your content KPIs will not otherwise capture. Building the non-branded query list you compare it against is keyword research work; a tool like Vrid.ai that generates the target query set alongside the content plan keeps the branded-vs-non-branded split consistent instead of stitched together from two different sources every reporting cycle.

Track this KPI monthly, not weekly. Branded search moves slower than ranking or citation metrics and is noisier at short intervals, so a weekly view mostly shows sampling noise rather than signal. It also compounds with the content architecture underneath it; sites that build topical authority through deliberate cluster architecture tend to see branded search grow faster than sites publishing the same volume with no internal-link structure tying it together.

KPI 4: Unlinked brand mention volume

Ahrefs analyzed 75,000 brands to find which off-site signals correlate most strongly with AI Overview inclusion. Brand web mentions, counting both linked and unlinked references, scored a Spearman correlation of 0.664 with AI Overview visibility. Backlinks scored 0.218 on the same scale, and brand search volume scored 0.392 (Ahrefs). The same study found that brands in the top quartile for web mentions earn up to 10x more AI Overview mentions than the next-closest quartile.

That ranks unlinked mentions above backlinks as a predictor of AI visibility, which reverses a decade of link-building priority. A mention with no hyperlink still teaches a language model to associate your brand name with a topic, a use case, or a category, because these systems are trained on the surrounding text, not on the link graph the way a PageRank-style crawler is. For the full case on why this reversal happened and how to measure it, see why brand mentions are becoming the new backlinks for AI search.

Track it with a brand-name search across mention-tracking tools (Ahrefs, Semrush, or a dedicated mention monitor), segmented into linked versus unlinked, and watch the unlinked share specifically. If your unlinked mentions cluster on a small number of domains, you have a concentration risk: losing access to one publication or community removes a disproportionate share of your AI-visibility signal. Reddit’s outsized share of AI citations, cited above in KPI 1, is exactly this kind of concentration, and it is worth reading the compliant playbook for building Reddit presence before you lean on it as a KPI driver, because getting flagged for astroturfing removes the signal entirely.

KPI 5: Assisted conversions

Assisted conversions count the touchpoints that happened before the final, credited touchpoint in a conversion path. In GA4, they live inside the Advertising section’s Conversion Attribution Analysis report, which separates single-touch from multi-touch paths and lets you compare how credit shifts across different attribution models (last-click, first-click, data-driven) for the same conversion set.

This KPI exists because SEO is one of the most consistently undervalued channels under last-click attribution: a visitor who reads three organic articles across two weeks and then converts from a branded paid search click gets zero credit assigned to organic under last-click, even though organic did the persuasion work. Pull assisted conversions and, if you have revenue or lead value connected to your GA4 key events, assisted revenue by channel, then compare organic’s assisted-conversion count to its last-click conversion count. A large gap between the two numbers is the size of the value last-click reporting was hiding.

Report this alongside, not instead of, last-click organic conversions. The point is not to inflate organic’s number by picking the attribution model that flatters it; it is to show the full conversion path so budget decisions do not defund a channel that is quietly doing assist work a last-click dashboard cannot see.

KPI 6: Non-branded organic conversion rate

With raw session volume no longer a reliable proxy for program health, the conversion rate of the sessions you do get becomes more informative than the session count itself. Segment out non-branded organic sessions specifically (excluding branded queries, which convert at a structurally higher rate because the visitor already decided), and track conversion rate on that segment month over month.

A falling non-branded conversion rate with flat traffic usually means the query mix shifted toward more informational, top-of-funnel intent, often because that is the intent still sending clicks post-AI-Overview while transactional queries get answered zero-click. A rising non-branded conversion rate with falling traffic can mean the opposite: you lost volume on low-intent queries and kept the visitors who actually convert, which is a healthier trade than the traffic number alone suggests. Neither read is possible from a sessions-only dashboard.

Segment by content type, not just by branded/non-branded. Comparison and calculator pages typically convert non-branded traffic at several times the rate of purely definitional content, so a blended conversion rate across a site’s full non-branded traffic obscures which page types are actually doing the work.

KPI 7: Content decay rate

Content decay rate measures the percentage of your published pages losing Search Console impressions over a trailing period (commonly 90 days), even before rankings visibly move. Impressions decay before position does, because Google’s continuous re-evaluation process narrows a page’s eligible query set gradually; by the time average position drops, the page has often already lost a meaningful share of the queries it used to be eligible for.

Pull it by comparing each page’s impression count for the trailing 90 days against the prior 90 days in Search Console, flagging any page down more than a threshold you set (a 20% impression drop is a reasonable trigger). Report the percentage of your indexed pages that cross that threshold each month, not the raw list, so the KPI is a single trend line you can put on a dashboard: rising decay rate across a growing share of your archive is the earliest warning that a site-wide relevance or freshness problem is developing, well before it shows up in aggregate traffic. For the decision framework on what to do with a decaying page, refreshing it or retiring it, see the refresh-versus-new-content framework and the pruning decision matrix by page class.

KPI 8: Cannibalization delta

Cannibalization delta tracks how many of your own pages compete for the same query in Search Console, and whether that number is growing or shrinking. Pull the query-level report, group by query, and count queries where two or more of your own URLs both show meaningful impressions. A rising count usually means your content program is publishing overlapping pages faster than your internal linking and canonicalization are resolving the overlap, which splits both ranking signal and, increasingly, AI citation signal, since an AI system choosing which of your pages to cite for a query is itself a form of cannibalization you do not control.

This metric matters more in 2026 than it did when every ranked page still got some click share regardless of position, because zero-click behavior means a page stuck at position 4 due to cannibalization from your own position-2 page is not just losing clicks to a competitor, it is losing them to nothing. Report the trend as a count of cannibalized query clusters, and pair it with the specific fix (consolidate, differentiate intent, or redirect) rather than a raw number with no action attached. The deeper query-level workflow for finding this, including decay and cannibalization together, is covered in the Search Console query analysis most teams never run.

KPI 9: Organic pipeline contribution

Organic pipeline contribution ties organic search, across both clicks and AI-driven brand awareness, to the sales pipeline it feeds: qualified leads sourced from organic, pipeline value attributed to organic touchpoints (using the assisted-conversion data from KPI 5), and closed revenue where organic appears anywhere in the conversion path. This is the metric that answers the question every budget conversation eventually asks, whether SEO drives the business, in a currency finance and sales leadership already use.

Building it requires your CRM and analytics to share a lead-source or UTM field that survives from first touch to closed deal, which most B2B teams already have for paid channels and frequently do not have wired correctly for organic. If organic leads land in the CRM tagged only as “website” with no channel breakdown, fix that plumbing before trying to report this KPI; a pipeline number built on bad attribution data is worse than no pipeline number, because it gets treated as authoritative in the room where budget gets cut.

Report it quarterly alongside the three-metric revenue view: organic-sourced pipeline, organic-assisted pipeline, and organic’s percentage of total pipeline by source. Quarterly, not monthly, because pipeline and close cycles for most B2B products run longer than a month, and a monthly view mostly shows noise from deal-timing variance rather than a real trend in organic’s contribution. If leadership asks you to project this forward instead of just reporting it backward, the scenario-based forecasting model with confidence bands is built for exactly that conversation, and pairs well with the worked ROI formula for a content program when the audience needs a break-even number.

Building the dashboard: one query set, three surfaces

The nine KPIs above only cohere if they are measured against the same underlying query set, tracked across the three surfaces where that query set now shows up: classic organic rankings, AI Overview and AI Mode, and third-party AI engines (ChatGPT, Perplexity, Gemini). Building three separate reports off three different query lists is how teams end up with dashboards that contradict each other for no real reason.

flowchart TD
    A[Define target query set:<br/>20-50 priority queries] --> B[Surface 1: Classic Google<br/>rankings, impressions, clicks]
    A --> C[Surface 2: AI Overviews/AI Mode<br/>impression share via GSC]
    A --> D[Surface 3: Third-party AI engines<br/>ChatGPT, Perplexity, Gemini]
    B --> E[Business layer:<br/>assisted conversions, non-branded<br/>conversion rate, pipeline contribution]
    C --> E
    D --> E
    E --> F{Quarterly review:<br/>which surface drives<br/>which outcome?}
    F --> G[Reallocate content and<br/>link-building effort by surface]

Start with the query set, not the tool. Pick 20-50 queries that map to real commercial intent, keep the list stable for at least a quarter so trend lines mean something, and run every KPI above against that same list. Add or drop queries only at the quarterly review, not mid-cycle, or you will not be able to tell a real trend from a query-list change. If you have not run a structured pass over your own site yet, the 40-check AI visibility audit is a faster starting point than building this dashboard from a blank page, and the end-to-end workflow from keyword to published post shows where these KPIs plug into a production pipeline rather than sitting in a separate reporting silo.

What to delete from your reporting deck

Total organic sessions without segmentation should come off the top-line dashboard. Reporting an aggregate session count with no split by query intent, content category, or conversion stage tells the room nothing about whether the program is working, since a million sessions from purely informational queries and a hundred thousand sessions from transactional queries can look identical on one line while representing completely different business value.

Aggregate Search Console impressions have the same problem, now made worse by AI Overview impressions inflating the total without any corresponding click data attached to explain why. Average keyword position across the whole site is similarly close to meaningless on its own: a #1 position on a query where you are not cited in the AI Overview can now perform worse, by clicks, than a #4 position on a query where you are cited, per the 35% CTR lift for cited brands in Seer’s data. Report position only alongside citation status for the same query, never alone.

None of this means stop collecting the old metrics. Keep them in the raw data layer for diagnosis. Stop putting them at the top of the executive-facing report, where they either overstate a decline that citation metrics would explain, or understate a real problem that a segmented view would catch earlier. If your reporting deck still leans on the metrics this section flags, the 11 AEO mistakes that keep sites out of AI answers usually explains why the citation-side numbers look worse than the ranking numbers suggest they should.

Frequently asked questions

What is the most important SEO KPI in 2026?

No single KPI works alone, because citation-side metrics (share of answer, AI Overview impression share) and business-side metrics (assisted conversions, pipeline contribution) answer different questions. If forced to pick one, organic pipeline contribution is the most defensible in budget conversations, but it depends on assisted-conversion and conversion-rate data underneath it to be trustworthy.

Should I stop tracking keyword rankings entirely?

No. Rankings still matter as a diagnostic input, especially paired with whether an AI Overview appears and whether you are cited in it. Stop reporting an average position number without that citation context, since a top ranking with no AI citation can lose 61% of its expected clicks on the same query (Search Engine Land).

How do I measure share of answer without a paid tool?

Run the same fixed set of 20-50 prompts manually against ChatGPT, Perplexity, and Google AI Overviews on a set schedule, log citations in a spreadsheet, and calculate the percentage by hand. It is slower than a dedicated visibility platform, but it produces the same underlying metric and works at any budget.

Does Google Search Console show AI Overview clicks yet?

No. The Generative AI performance report Google launched in June 2026 shows impressions from AI Overviews and AI Mode only, with no click, CTR, position, or query-level data attached, and it started rolling out to a subset of UK sites first (ppc.land). Track impression share as the proxy metric until Google adds click data.

Why did my traffic drop with no announced Google update?

Google’s continuous re-evaluation of quality and relevance runs outside announced core update windows, and third-party volatility trackers regularly record elevated ranking movement in the gaps between confirmed updates. Check content decay rate (KPI 7) first, since impression loss usually precedes a visible ranking drop by weeks.

Are backlinks still worth building in 2026?

Backlinks still carry ranking weight and referral traffic, but Ahrefs’ study of 75,000 brands found they correlate with AI Overview visibility at 0.218, well below unlinked brand mentions at 0.664 (Ahrefs). Treat mentions and links as separate KPIs with separate acquisition tactics rather than one combined “authority” number.

What counts as a good non-branded conversion rate benchmark?

There is no universal benchmark; it depends heavily on content type, industry, and funnel stage. Track your own non-branded conversion rate as a trend against its own history, segmented by page type, rather than comparing it to an external average that mixes content types you do not publish.

How often should I run a share-of-answer check?

Weekly or biweekly for your top 20-50 priority queries. AI engines re-crawl and update their source sets faster than classic Google rankings refresh, so a monthly cadence misses meaningful movement, especially around product launches or news cycles relevant to your category.

Should branded and non-branded search be reported together?

No. Branded search converts at a structurally different rate because the visitor already knows your brand, and blending it with non-branded traffic hides whether your top-of-funnel content is actually attracting new demand or simply capturing people who were already looking for you by name.

What is content decay rate and why does it matter more now?

It is the percentage of your published pages losing Search Console impressions over a trailing period, typically 90 days. It matters more in 2026 because impression loss is now the earliest available warning sign, arriving weeks before a visible position drop, in an environment where you cannot afford to wait for the lagging signal.

How do I attribute revenue to SEO when most searches don’t result in a click?

Use assisted conversions and multi-touch attribution in GA4 to capture the touchpoints organic contributes before a different channel gets last-click credit, and pair that with branded search volume growth and unlinked mention volume as proxies for the awareness organic generates outside any tracked session at all.

Is share of voice dead?

Not dead, but incomplete. Share of voice still measures presence in classic rankings and should stay in the diagnostic layer of your reporting. Share of answer measures a related but distinct thing, presence inside AI-generated answers, and the two numbers frequently diverge for the same query set.

What tools track AI citation and share of answer?

Dedicated AI-visibility platforms exist for this, alongside the manual-prompt method described above. Regardless of tool, the requirement is the same: a fixed query set, a repeatable prompt structure, and a consistent logging cadence, since the metric is only useful as a trend, not a single snapshot.

How do I know if cannibalization is hurting my AI citations, not just my rankings?

Check which of your own URLs an AI engine cites for a query where you have multiple competing pages. If the engine cites your weaker page, or alternates between pages across repeated checks, that is a cannibalization signal specific to AI citation, separate from whatever your classic Search Console query report shows.

Do I need llms.txt to improve these KPIs?

Not for Google. Google’s Gary Illyes confirmed in July 2025 that Google Search does not use llms.txt, comparing it to the old keywords meta tag, though Perplexity and Claude do consume it according to reporting from Baseline Labs. It will not move Google-surface KPIs, but it may affect citation behavior on the engines that read it.

How do I report these KPIs to executives who only understand traffic?

Lead with organic pipeline contribution and assisted conversions, since those translate directly into revenue language leadership already tracks for other channels. Use share of answer and AI Overview impression share as supporting evidence for why traffic-only numbers look flat or declining, not as the headline metric itself.

Should ecommerce sites track these KPIs differently?

The framework holds, but weight KPI 6 (non-branded conversion rate) and KPI 9 (pipeline contribution, adapted to direct revenue) more heavily than share of answer for bottom-funnel transactional queries, since AI Overview citation matters less on queries with 31% zero-click rates than on the 74% zero-click informational queries where citation drives most of the remaining visibility.

How long before a new KPI dashboard shows a usable trend?

Give it one full quarter minimum before drawing conclusions, and two quarters before making budget decisions off it. Several of these metrics, especially branded search volume and pipeline contribution, are intentionally slow-moving and noisy at shorter intervals.

What is the single biggest reporting mistake teams make in 2026?

Keeping a sessions-first dashboard unchanged while adding one or two AI-visibility metrics on the side, so the top-line story still reads as decline even when citation and pipeline metrics show the program working. Restructure the hierarchy, not just the metric list, so the metrics that predict revenue lead the report.

Where do I start if I have never tracked citation metrics before?

Pick your 20 highest-intent, highest-value queries, the ones closest to a purchase decision, and run manual share-of-answer checks on those first rather than trying to build a full 50-query system on day one. Layer in AI Overview impression share from Search Console next, since that data already exists and requires no new manual process.

Key takeaways

Sessions and blended CTR describe less of your program every quarter, because 68.01% of searches now end without a click and AI Overview queries push that closer to 83% (Search Engine Land). Replace a clicks-only view with citation-side metrics (share of answer, AI Overview impression share, branded search growth, unlinked mentions) and business-side metrics (assisted conversions, non-branded conversion rate, content decay rate, cannibalization delta, pipeline contribution), measured against one stable query set across all three search surfaces.

Building and maintaining that query set is keyword research work that has to stay current as query behavior shifts between classic search and AI engines; Vrid.ai handles keyword research alongside AI-assisted content production so the same query list drives both your KPI tracking and what you publish next, instead of the two drifting apart every reporting cycle.