The First-Week SEO Audit for a New Client, Using Only Search Console Data
A first-week SEO audit for a new client using only Search Console: what to confirm first, seven checks and a one-page prioritized plan.
A good SEO audit for a new client fits on one page. Before you read any number, confirm the property type, your permission level, the data delay and the 16-month history limit. Then run seven Search Console checks: a clean traffic baseline, brand vs non-brand, pages losing clicks, positions 4-20, cannibalization, indexing and origin-level Core Web Vitals. Score each finding by impact and effort, and promise measurement, not rankings.
The first audit should produce a decision, not a document
A new client has just added your Google account to their Search Console property. They want to know what you will work on first, and they are not going to read a 60-page PDF to find out.
So the goal of week one is narrow: a short list of problems and opportunities, each backed by a number you can show, ranked by how much it matters and how hard it is to fix. Everything below uses Search Console alone, the one data source almost every client already has.
Each check has two routes: the free Search Console interface, or a one-sentence prompt for an AI assistant connected through GSC Magic that calls the matching report tool. For the general AI audit method, see the AI SEO audit guide; this article covers the onboarding version and how to prioritize it.
Confirm four things before you look at a single number
Most bad first audits come from reading the wrong data correctly. Spend the first hour here.
Property type. A Domain property covers every subdomain and both http and https. A URL-prefix property covers only URLs that start with that exact prefix, protocol included. Google’s property type documentation gives the example that https://example.com/pets/ excludes https://m.example.com/pets/. If the client gave you a URL-prefix property for https://www., you are not seeing the blog on blog. or the old http:// pages. Ask whether a Domain property exists before you call a number “site-wide”.
Your permission level. A restricted user has view rights on most data, but per Google’s permissions table gets view-only access to indexing reports and cannot submit sitemaps. That is fine for an audit. It is not fine for fixing things, so ask for the right level now rather than in week three.
The data delay. Google’s Analytics help states that Search Console data becomes available about 48 hours after it is collected, and the Performance report marks the newest data as preliminary. A safe rule of thumb: end every comparison window three days before today, so a half-collected day never looks like a drop.
The 16-month limit. Search Console keeps the last 16 months, per the same Google page. A collapse two years ago is invisible to you. Ask whether anyone exported older data, and read how to keep data longer than 16 months if they want history to survive this time.
Also note that interface tables export at most 1,000 rows, so on a large site the manual route shows only the top of each list. The 1,000-row limit guide covers workarounds.
Check 1: a traffic baseline with two equal, complete windows
The question: is Google Search traffic growing, flat or falling, and is the change in demand (impressions) or in clicks per impression (CTR)?
By hand: open Performance, then Search results, set Search type to Web and use Date, then Compare, with two custom ranges of equal length. Last 28 complete days against the 28 before is a good default. Then switch to the full 16 months to see the shape of the year, and compare with the same 28 days a year earlier if the business is seasonal.
Ask the assistant:
Using gsc_query_analytics with the date dimension, compare clicks, impressions, CTR and average position for the last 28 complete days against the previous 28, ending three days ago, and show the daily trend for the last 16 months.
What bad looks like: impressions steady while clicks fall points to CTR (a SERP feature, a title change, a position slide from 2 to 5). Impressions falling with them points to lost visibility or lost demand. If the year-over-year line explains the movement, write “seasonal” and move on; the seasonality guide helps you tell the two apart.
Write the baseline into the plan with its exact dates. Month three gets compared against it.
Check 2: brand vs non-brand tells you how much of the traffic is yours to grow
The question: what share of clicks comes from people who already know the company name?
By hand: the Performance report now has a branded queries filter. Its documentation says it is not available for sites with a low number of impressions and that some queries may be classified wrongly. Where it is missing, use the Query filter with “queries containing” the brand name and its common misspellings, then repeat with “queries not containing”.
A detail that trips people up: per the same page, anonymized queries drop out of the chart totals once a query filter is applied, so branded plus non-branded clicks add up to less than the unfiltered total. Report shares, not a sum that contradicts the headline number.
Ask the assistant:
Run report_brand_split for this property with the brand terms “acme”, “acme tools” and “acmetools”, and give me clicks, impressions and click share for each group plus the top queries in both.
What bad looks like: a site where most clicks are branded has little non-brand visibility, so “SEO traffic” is mostly brand demand that marketing or sales created. That is not a failure, but it changes what you can promise. More detail is in brand vs non-brand in Search Console.
Check 3: find the pages that explain most of the lost clicks
The question: which URLs lost the most clicks between the two baseline windows, in absolute numbers?
By hand: keep the comparison from check 1, open the Pages tab and sort by the click difference. A page going from 900 to 600 clicks matters more than one going from 6 to 0, even though the second is a 100% drop. Click the top losers and look at their queries in the same comparison.
Ask the assistant:
Run report_traffic_drops by page for the last 28 complete days against the previous 28, then run report_content_decay and tell me which losing pages also fell in each of the last three windows.
What bad looks like: two or three URLs accounting for most of the site’s loss, especially a commercial page. A page that shows up in both reports is sliding steadily. Outdated content or a stronger competitor are the usual suspects, but treat any cause as a hypothesis until you have checked the live page. The traffic drop checklist is the deeper version of this step.
Check 4: striking-distance queries are the cheapest growth on the list
The question: which query and page pairs already rank in positions 4 to 20 with enough impressions to matter?
By hand: in Performance, open Queries, add the Position column, filter the table to positions above 3 and below 21 and sort by impressions. Click a query, then the Pages tab, to see which URL ranks. Positions 4-20 is a common working range, not a Google definition.
Ask the assistant:
Run report_quick_wins for this property over the last 28 complete days and list the top 15 query and page pairs by potential extra clicks, with position and impressions.
What bad looks like: the strongest opportunities mapped to the wrong page (a blog post ranking for a query the product page should own), or a list made entirely of queries no customer would type. Filter for relevance before you put anything in the plan. The full method is in striking-distance keywords in Search Console.
Check 5: cannibalization shows up as two URLs sharing one query
The question: are two or more of the client’s pages each getting a real share of impressions for the same query?
By hand: take the top 20 non-brand queries from check 2. For each, filter Performance to that exact query and open the Pages tab. Two URLs that both collect meaningful impressions, with neither in the top three, is the pattern to note.
Ask the assistant:
Run report_cannibalization for this property and show the queries where two or more pages split impressions, with each page’s share of impressions.
What bad looks like: a category page and a blog post trading places for a commercial query, both stuck mid-page. A homepage and a product page sharing a brand query is normal, so flag only overlaps where both pages target the same intent. The cannibalization prompt guide shows how to decide which page should win.
Check 6: indexing problems hide in the sitemap gap
The question: are the pages the client cares about indexed, and how far apart are “submitted” and “indexed”?
By hand: open Indexing, then Pages, and switch the dropdown above the chart from “All known pages” to “All submitted pages” to limit the report to sitemap URLs. In the Sitemaps report, check each status (Success, Has errors or Couldn’t fetch) and the Discovered pages count; Google’s Sitemaps report help notes that a discovered URL is not guaranteed to be indexed.
Then read the reasons. The Page indexing report documentation lists them by name. On a first audit, these deserve attention when they hit submitted, important URLs:
- “URL marked ‘noindex’” on a page that should rank
- “Duplicate, Google chose different canonical than user”
- “Crawled - currently not indexed” on commercial pages
- “Server error (5xx)” or “Soft 404”
Ask the assistant:
Run report_index_health for this property, group the sitemap URLs by coverage state, and list every submitted URL that is not indexed with its coverage state.
What bad looks like: a sitemap full of redirects, noindexed URLs or 404s, which wastes the signal, or a revenue page excluded by a stray noindex. Many excluded URLs are excluded on purpose (“Page with redirect” and “Alternate page with proper canonical tag” are often correct). The sitemap vs indexed guide explains which gaps to chase.
Check 7: read Core Web Vitals at origin level before you open Lighthouse
The question: do real visitors get a good experience across the site, on mobile and desktop?
By hand: open the Core Web Vitals report under Experience. Its documentation says the data comes from the Chrome UX Report, that URLs are grouped by similar experience and that a group’s status is set by its worst metric at the 75th percentile. Good means LCP up to 2.5 s, INP up to 200 ms and CLS up to 0.1. Low-traffic sites may have no data at all.
Ask the assistant:
Run cwv_get_origin for the client’s origin, then cwv_history for the same origin, and tell me which metric is worst and whether it improved or got worse over the period.
What bad looks like: a Poor INP or LCP at origin level, especially on mobile. In a first audit, Core Web Vitals work as a gate. A Poor origin goes on the plan. Good or Needs improvement rarely beats a noindexed money page or a page losing 300 clicks a month.
Turn the seven checks into a one-page plan
By Friday you should have between ten and twenty findings. The client needs five. Score each finding on two axes:
- Impact: how many clicks or how much business value is plausibly at stake. Use the numbers from the checks (clicks lost, impressions in positions 4-20, revenue pages affected), not adjectives.
- Effort: who has to do the work and how long it takes. A noindex tag removed by the developer in ten minutes is low effort. Merging two cannibalizing articles with redirects is medium. Rebuilding a template for INP is high.
Then sort into three groups:
| Group | What goes in it | Example |
|---|---|---|
| Fix this month | High impact, low effort | Pricing page marked noindex |
| Plan next | High impact, higher effort | Two category pages splitting one commercial query |
| Park | Low impact, any effort | Old news post losing 15 clicks |
Each item on the page gets one line of evidence (the URL, the metric and the dates), one line of proposed work and one line saying how you will measure it. That is the whole document. If you want to turn it into a priced scope of work, the proposal guide picks up from here.
What not to promise after a first audit
Search Console describes the past, so the plan should not predict Google’s next move.
Do not promise rankings. Do not promise a number of extra clicks by a date, even when report_quick_wins shows a potential figure, because that estimate assumes a position change you do not control. Do not promise that every “not indexed” URL will become indexed, because many should stay out. Do not promise that better Core Web Vitals will lift traffic by itself.
What you can promise: which pages you will change, when, and which Search Console numbers you will compare afterwards over equal, complete windows. That one you can keep.
Frequently asked questions
How long should a first SEO audit for a new client take?
For a small or mid-sized site, about a week of part-time work: one day to confirm the property and data limits, two or three days for the seven checks, one day to verify findings on the live site and one day to write the plan. A larger site needs more time for indexing and cannibalization review.
Is Search Console data enough for a new client SEO audit?
It is enough for a first-week audit focused on Google Search performance, indexing and real-user Core Web Vitals. It does not show backlinks, competitors, conversions or the content of the pages, so any finding that depends on those needs a separate check before you act on it.
Should the first audit include a traffic forecast?
No. Search Console shows what already happened. Use it to estimate which work could matter most, and describe the expected effect as a direction to measure, not a number of clicks by a date.
Repeat the audit for every new client without rebuilding it
For one new client a quarter, the manual route is all you need. By the fifth client, the repeated filters and exports start to cost real hours.
GSC Magic connects Claude, ChatGPT and other MCP clients to Search Console, Bing Webmaster Tools and GA4, so the seven prompts above run against the client’s real property. Every new account starts with a 7-day Pro trial without a card. Pro covers up to 10 sites for $19 a month (a founding price until the end of 2026), and the Agency plan raises that to 25 sites with five seats and client groups so each person and each agent sees only their client’s data. Plan details are on the pricing page.
Start with the property you onboarded this week. Run check 1, write the baseline down with its dates, and build the rest of the plan on top of it.
Give your agent your search data.
GSC Magic connects Claude, ChatGPT and Cursor to Google Search Console, Bing and GA4. Free plan, 7-day Pro trial, no card.