When a SaaS SEO agency's pitch is "we read your Search Console first": what a buyer should be able to query themselves before paying for the reading
A buyer's pre-read of Search Console, rendering, and AI access.

A SaaS SEO agency’s offer to read Search Console is a starting point. Before paying for that reading, a technically literate buyer can inspect the main Google search signals, check access and rendering, and build a small AI visibility baseline. That pre-read turns the agency call into a test of explanation.
Why the "we'll read your Search Console" agency pitch is not a diagnosis
The sales ritual borrows authority from the instrument. Search Console contains real, private data, so asking for access sounds more rigorous than arriving with a generic pitch. But access is not analysis, and an issue count does not explain what prevents a revenue page from reaching buyers.
A useful first question is what the agency expects to find from what it can already see outside the account. Its answer should contain a hypothesis. If the proposed first weeks begin with a content calendar, has the agency considered whether the constraint is technical? The buyer’s job before that conversation is not to finish the audit, but to establish enough evidence to judge the answer.
What Search Console actually shows and where it stops
Search Console records what Google has observed. It does not record what buyers ask ChatGPT or Claude, and it supplies no first-party prompt data from either. Google includes traffic from AI Overview and AI Mode in the normal Web search type in Search Console. Those clicks are not a separate AI channel that a buyer can isolate from ordinary organic clicks in Console alone.
Even within Google, the query report needs interpretation. It presents raw search queries that require interpretation to distinguish keywords from AI prompts. Question-based, context-rich and personalized queries need to be read differently from short head terms because they can show different buying needs. A problem question, an alternatives search and an implementation question belong to different stages of a decision. Classification helps reveal those stages, but it does not turn the report into a window onto other assistants.
The second channel Search Console cannot see: AI crawler access
To check the access layer, open the site’s robots.txt and look for Disallow rules targeting GPTBot, ClaudeBot or PerplexityBot. This is a quick public check. A block identifies an access obstacle that better writing alone cannot fix.
That separation gives the buyer an early evaluation point. If a block is present, the agency could have spotted it from outside the account. Did it mention the block unprompted, or discuss AI citations without checking whether these crawlers could reach the site? The omission gives the buyer a concrete gap to probe.
Seven failure classes a real audit maps — and how to spot which one you have before the call
A useful audit sorts observations into seven failure classes, namely missing page, relevance, authority, indexation, technical access, conversion and measurement. For the pre-read, work through them in that order against the site’s most commercially important pages. The pre-read should produce a map of likely causes.
Start with missing pages. Check whether the site has destinations for searches such as “[competitor] alternatives”, “[your brand] vs [competitor]”, “[category] software for [industry]” and “[your tool] + [integration]”. These queries come from buyers comparing tools or checking fit. An agency that proposes more content should be able to name the page buyers need that the site lacks.
Next, check relevance using the queries already in Search Console. Does the page that ranks help the reader compare tools or check fit, or send the reader into a blog post about the general problem? A page can exist and still miss the intent. The agency should distinguish that mismatch from a genuinely missing page.
For authority, check the internal links to revenue pages. Do high-traffic pages link to the relevant integrations and comparisons, or are those destinations orphaned? This is free to check before the call. An agency that notices weak revenue pages should also examine how the rest of the site connects buyers to them.
For indexation, use URL Inspection on the five most commercially important pages and look for “Crawled, currently not indexed”. Then examine likely sources of confusion, including inconsistent canonicals, documentation parameters, crawlable staging or legacy subdomains, and duplicate versioned docs. Expired feature pages that return 200 with an error message belong here too. Google treats rel="canonical" as a strong signal, while robots.txt controls crawler access. An agency should keep those mechanisms distinct.
Technical access needs a different check. Compare the raw HTML of a pricing or feature URL with the rendered page. If key content appears only after JavaScript executes, the first response may be little more than a shell. The agency should recognize that this content depends on rendering.
For conversion, check whether the revenue page has a clear next step that works without JavaScript. A trial or demo CTA that appears only after a script fires creates both an access concern and a conversion concern. Those are related, but not interchangeable. The agency’s diagnosis should explain both concerns as effects of the same symptom.
Finally, check what the measurement covers before trusting the totals. Confirm that Search Console has verified coverage for the domain and both the www and non-www variants. Partial coverage makes the reading partial too. An agency should establish which property it is reading before presenting conclusions about the whole site.
What the Core Web Vitals report tells a buyer in five minutes
For a five-minute check, open Core Web Vitals in Search Console, filter to the Poor URLs list, and cross-reference it against pricing, feature, comparison and integration pages. Note the affected revenue URLs and the reported LCP, INP or CLS failures. No deep performance tutorial is needed for this step.
An agency can check the pricing page’s public PageSpeed Insights score before it has access to the account. A technically literate team should be able to connect likely failures to the site’s React or Next.js architecture. Naming a poor result is easy enough. Explaining its shape is where the diagnostic work begins.
The rendering gap that makes SaaS indexation different from brochure-site SEO
A SaaS site can show users a rich page while supplying much less content in its initial HTML response. React and Angular single-page applications can produce this gap, while Next.js is designed to support server-rendered HTML on the first request. The framework name alone is not the diagnosis. The diagnosis depends on how each revenue page actually delivers its content.
To inspect that difference, use browser developer tools or a “Fetch as Google” equivalent to compare the initial response with the rendered DOM. Look for pricing tiers, feature bullets, customer logos and comparison tables in particular. Dynamically loaded content, modal-based tours and API-delivered testimonials may appear only after rendering. If they are missing from the initial HTML, there is an indexation risk, even when a standard crawl reports no errors.
The agency’s answer should therefore discuss SSR, CSR or ISR for the relevant page types. That is the technical depth the call needs. The question to ask is “Given our rendering stack, which of our revenue pages do you expect to find in the render queue, and how would you verify?”
How to run a manual AI visibility pre-read before the call
The AI channel needs its own baseline. A manual pre-read that a buyer can trust takes under an hour and needs no paid tool. First, build a prompt set from sales call notes, support tickets, demo form fields, Search Console queries and community threads. Cover problems, categories, competitors and alternatives, integrations, pricing and implementation.
Next, run each prompt in ChatGPT with search enabled, Perplexity and Google, checking Google for AI Overviews. Use a logged-out or clean browser session for each. Record the date, platform and exact prompt, then note whether the brand appears, which URL receives a citation and which competitors appear. Keep a separate record of third-party sources, including review sites, publications and forums. G2, Capterra, Reddit and YouTube are worth tracking because they show where the category conversation lives.
For each absent brand result, investigate the underlying page. Does a relevant page exist, and is it indexed? Does it contain criteria, data, limitations and dates, or only claims? Is it connected to other pages on the topic?
Finally, filter GA4 referral data for sessions whose session source contains “perplexity.ai”, “chatgpt.com”, “claude.ai” or “you.com”. This adds a downstream signal of citations that drove visits. The prompt log shows the answers, while referral data shows the visits. With both in hand, the buyer can compare the agency’s findings against an existing baseline.
What the agency's response to your pre-read reveals about its methodology
It helps when the agency agrees with what the buyer found, but treating those same findings as new adds little. What does the agency explain that the buyer cannot yet explain? The agency should show its value by linking a symptom to its likely cause and what it plans to do about it.
Public checks provide the first test. Did the agency get the rendering architecture right and check the AI crawler status before asking for access? Funnel questions provide another. Did it ask about trial length, average time-to-convert and the definition of a qualified demo before recommending keywords? A content strategy should then change by funnel stage, with named examples.
The AI proposal deserves equal scrutiny. Ask how the agency expects to earn citations for buying-intent queries and which platforms it tracks. At the audit stage, give it read access to Search Console only. Establish ownership before the work begins too. The answer about ending the engagement should be immediate and clear, with the buyer owning everything and transition happening on the day of notice.
What good AI visibility coverage looks like — and how to tell if an agency actually measures it
A single AI visibility score blends surfaces that cite different sources. Per-platform tracking is the minimum useful commitment, covering at least ChatGPT, Perplexity, Gemini and Google AI Overviews/AI Mode. The same prompts should run on the same platforms at a fixed cadence. Otherwise, a change in results is harder to distinguish from a change in the test.
Citation sources belong in that record too. An agency should track which third-party pages each platform cites for the category alongside appearances of the brand’s own pages. GA4 referral data then adds evidence of visits generated by those citations. Answer logging and referral measurement serve different purposes, so neither should stand in for the other.
Tools can automate parts of this work. Promptwatch states that it tracks brand mentions across twenty AI models and captures answers from real consumer interfaces as well as APIs. CrunchJunkie, made by Tiki-Taka Media GmbH in Hamburg, tracks mentions and source citations across ten AI engines on every plan, including its entry tier. Coverage is not a methodology by itself. The buyer’s existing prompt log and referral filter make it possible to ask exactly what the agency’s measurement adds.
The content properties that make a page citation-eligible across both channels
After diagnosis comes the question of what deserves publication. A specific claim, verifiable data, a named limitation or a first-hand example gives an answer engine something worth extracting, whether or not a model wrote the draft. For a buyer, that same detail supplies grounds for a decision.
Four properties make the content program concrete enough to evaluate. A page should answer a specific buying question with criteria, dates, limitations and the logic behind the decision. It should be indexed and reachable without authentication or JavaScript execution. What it says about the brand should match what the site, LinkedIn, Crunchbase, review platforms and third-party mentions say. It should also connect to other pages on the same topic.
These properties turn a content proposal into something testable. Which existing pages pass all four, and which proposed deliverables repair a specific failure? A buyer who can ask that question is no longer paying merely to hear their own Search Console read aloud. They are evaluating whether the agency can turn evidence into a defensible plan.


