AI Overviews, ChatGPT, and Perplexity have added a new type of traffic problem forensic SEO has to account for: pages that still rank well, or are even cited as sources, but no longer get clicked because the answer is delivered directly on the results page or inside a chat response. Forensic SEO in the AI search era means learning to tell that apart from an actual ranking loss — because the two look identical in a basic traffic graph and require completely different remediation.
- Why AI Search Changes the Forensic Question
- How to Tell Ranking Loss From Click Absorption
- Diagnosing Whether Content Is Actually Being Used by AI Systems
- New Failure Modes Unique to AI Search
- Building an AI Visibility Baseline Before You Need One
- What Remediation Looks Like for AI-Era Findings
- Keeping Traditional Forensic Rigor Intact
Why AI Search Changes the Forensic Question
Traditional forensic SEO asks: did this page stop ranking, and why? That question still matters, but it’s no longer the whole picture. A page can hold its position 1 ranking and still lose most of its click volume because an AI Overview now answers the query directly above it, or because a chat assistant answered the user’s question without the user ever visiting a search engine at all. The symptom — falling organic sessions — looks the same in Google Analytics whether the cause is a ranking drop or a click absorption problem. The forensic work now has to separate them before any remediation plan makes sense.
This distinction isn’t academic. Fixing a “ranking loss” that’s actually a click absorption problem means rewriting content, building links, or restructuring pages that were never actually broken — the fix will not move the metric that’s actually declining, because rankings were never the problem.
How to Tell Ranking Loss From Click Absorption
The diagnostic move is straightforward once you know to look for it: pull query-level Search Console data and compare impressions against clicks separately, rather than looking at clicks alone.
- Impressions falling, clicks falling proportionally: consistent with an actual ranking loss — the page is being shown less often.
- Impressions stable or rising, clicks falling: consistent with click absorption — the page is still being shown (and possibly still ranking well) but users aren’t clicking through, often because an AI Overview or featured snippet is answering the query above the organic results.
- Average position holding steady while both metrics fall: a strong indicator the SERP itself has changed structurally — more ads, more AI-generated content, or a new SERP feature pushing the traditional result further down the page even without a ranking change.
Manual verification closes the loop: search the affected queries directly (logged out, incognito, matching the target geography) and check whether an AI Overview now appears where it didn’t before. Google doesn’t expose AI Overview presence historically in Search Console, so this manual sampling — done consistently on a defined query list — is currently the most reliable way to build a before-and-after record.
Diagnosing Whether Content Is Actually Being Used by AI Systems
A separate forensic question in the AI era: is the site’s content actually being cited or drawn from by AI systems, even without producing a click? This matters for brand visibility even when it doesn’t move traffic directly.
Perplexity is the most useful tool for this check because it displays explicit source citations for every answer, making it possible to directly observe whether a site is being surfaced for a given query and how consistently. Running a defined set of target queries through Perplexity, ChatGPT, and Google AI Overviews on a recurring schedule — and logging whether the site is cited, paraphrased without attribution, or absent — builds the kind of longitudinal evidence a forensic investigation needs, since none of the major platforms currently provide this history natively.
New Failure Modes Unique to AI Search
Beyond click absorption, AI search introduces failure modes that didn’t exist in a purely traditional-SERP world:
Content being summarized without attribution
A page can be the effective source of an AI-generated answer while receiving zero citation or click — the content gets used, the brand gets nothing. This is difficult to prove definitively but worth documenting when a distinctive phrase, data point, or framework from a site’s content appears in an AI answer with no attribution.
Entity confusion suppressing citations
Sites with weak or inconsistent entity signals — inconsistent business name formatting, no structured author or organization data, conflicting information across the web — can be deprioritized by AI systems that favor sources they can confidently identify and vet. This is a forensic angle worth checking on sites that rank reasonably well but never seem to get cited in AI answers: is the entity itself clearly and consistently defined across the site and the wider web?
Structured data gaps
Missing or broken Article, FAQPage, Organization, or Person schema doesn’t just miss out on rich results — it removes a machine-readable signal that both traditional crawlers and AI retrieval systems use to quickly understand what a page is about and who’s accountable for it. A forensic check worth running: validate schema on pages that seem to be losing AI visibility, since a broken or missing schema implementation is a common, fixable contributor.
Building an AI Visibility Baseline Before You Need One
The same principle that applies to traditional forensic SEO — you need a “before” snapshot to diagnose a “what changed” question — applies to AI search, and almost no one has one yet, because AI Overview and chat-based citation tracking is still a manual, recent practice for most sites. The practical fix: start logging AI visibility for a defined set of priority queries now, even if nothing looks wrong today. A simple spreadsheet — query, date, platform, cited or not, position of citation if visible — becomes genuinely valuable evidence the first time a client asks why their AI visibility seems to have changed.
At Salterra, this baseline-building has become a standard addition to ongoing client reporting, specifically because we’ve already run forensic cases where the client’s honest answer to “were you cited in AI Overviews before?” was “we have no idea,” which made the investigation far slower than it needed to be.
What Remediation Looks Like for AI-Era Findings
Remediation depends entirely on which failure mode the evidence points to, and it’s rarely the same fix as a traditional ranking problem:
- Click absorption from AI Overviews: restructure content to answer the core question immediately and completely, then go deeper — the goal shifts partly from “win the click” to “win the citation and capture the clicks that do happen from users wanting more depth.”
- Missing AI citations despite strong content: strengthen entity signals — consistent NAP and brand data, named author bios with verifiable credentials, Organization and Person schema, and off-site presence that reinforces the same entity information.
- Uncited summarization: there’s no reliable technical fix here yet; the most durable response is building enough off-site brand presence and citation history that being the recognized original source becomes harder for a model to ignore over time.
Keeping Traditional Forensic Rigor Intact
None of this replaces the traditional forensic SEO workflow — it adds a layer to it. A drop that looks like AI click absorption still needs to be checked against the standard timeline of site changes, algorithm updates, and technical issues, because it’s entirely possible for a real ranking problem and a real AI Overview presence to be happening on the same page at the same time. Treating “it’s probably the AI Overview” as an automatic excuse is its own version of the mistake of accepting the first plausible explanation without testing it against the full body of evidence.
Frequently Asked Questions
How do I know if an AI Overview is stealing my clicks?
Compare Search Console impressions and clicks at the query level. If impressions are stable or rising while clicks fall, and a manual search of the query shows an AI Overview now appearing above the organic results, that's a strong indicator. This is different from a ranking loss, where both impressions and clicks typically fall together.
Can I recover traffic lost to an AI Overview?
Not always to the same volume, since some users will always be satisfied by the on-page answer and never click through regardless of content quality. But you can improve the odds of being the cited source (capturing brand visibility even without a click) and improve the click-through rate for users who do want more depth, by making sure the content clearly goes beyond what the AI Overview itself can summarize.
Do I need special tools to track AI search visibility?
Not necessarily. Manual sampling in Google, ChatGPT, and Perplexity against a defined query list, logged consistently, is a legitimate and currently common approach since dedicated tracking tools are still maturing. Some rank tracking platforms are adding AI Overview presence tracking, which is worth using if available, but it should supplement manual checks, not replace them entirely.
Is it worth optimizing for AI citations if my business relies on click-through traffic?
Yes, for most informational and educational content, because AI visibility and traditional ranking strength share the same underlying foundation — clear structure, direct answers, strong entity signals, and genuine expertise. Work done to improve AI citation likelihood rarely hurts traditional SEO performance and often improves it, since both reward the same underlying content quality.
How often should I check AI search visibility for forensic purposes?
A monthly manual check against a defined set of priority queries is a reasonable baseline for most sites, increasing to weekly for sites in categories where AI Overviews appear frequently or where a client has specifically flagged AI visibility as a priority. The exact cadence matters less than doing it consistently enough to have a real historical record when a forensic question eventually comes up.
Terry has 30+ years in software and SEO. He’s the founder of Salterra Digital Services and SEO Spring Training, host of the Roundtable SEO Mastermind, and lead instructor at SEO University — teaching the exact tactics his team uses on client work.
Continue learning
Ready to master this?
This guide is one lesson from the Forensic SEO course. Get every lesson, framework and checklist — plus the full 38-course catalog — inside SEO University.
Practitioner-focused training across the full digital marketing stack — from technical SEO to conversion optimization and the AI search era. By Salterra Digital Services, since 2011.