How to Write an SEO Proposal Backed by the Prospect's Own Search Console Data

Write an SEO proposal from 30 minutes of the prospect's Search Console data: how to ask for access, five findings, an honest estimate and a copyable outline.

TL;DR

The strongest SEO proposal is built on 30 minutes of the prospect’s real Search Console data, not on a template and a tool score. Ask for temporary Restricted access, pull five findings (queries at positions 4-20 with few clicks, pages that lost clicks between two equal windows, cannibalized queries, brand vs non-brand share, indexing gaps), turn them into a ranged estimate with stated assumptions, and never promise rankings. If they won’t share access, pitch from public data and say clearly what you can’t know. If they don’t sign, remove yourself.

Most SEO proposals lose because they are about you

A generic SEO proposal template has the same skeleton everywhere: an agency intro, a screenshot of a site-health score, a list of services and three package prices. The prospect has seen it from every other freelancer they talked to. Nothing in it proves you looked at their site.

My position is simple. A proposal built on half an hour of the prospect’s own Search Console data beats a polished template almost every time. It names their pages. It quotes their queries. It contains an estimate they can check against their own account. That changes the conversation from “do we trust this person’s promises” to “is this person right about our numbers”.

It also protects you. A proposal anchored in real data is harder to haggle down to a checklist, because the scope comes from findings, not from a menu.

This guide covers the pre-sale part: getting access, what to pull, how to estimate, how to structure the document. The full audit after signing is a separate job, covered in our new client audit guide.

Ask for Restricted access, in writing, with an end date

Ask for Restricted user access to their Search Console property, for a fixed period, before you write the proposal. Restricted is the lowest login role Google offers and is enough for everything in this guide.

Google’s permissions page lists the levels: owner, full user and restricted user. A restricted user “has simple view rights on most data”: performance data, most reports, links. A restricted user cannot add users or owners, cannot submit sitemaps and cannot disavow links. Full users get more actions, and you don’t need any of them to write a proposal.

The owner adds you under Settings, Users and permissions, Add user, using your Google account e-mail. Only owners see that screen, so if your contact is a marketing manager without ownership, they need to forward the request to whoever owns the property. For the click-by-click version to send them, see our guide on giving an SEO agency access to Search Console.

The request itself matters more than people think. Say what you will look at, what you won’t do and when access ends. Here is a message you can copy:

Hi [name],

Before I send a proposal, I'd like to base it on your real search data
instead of a generic template. Could you add me to Google Search Console
as a Restricted user (view only)?

Search Console > Settings > Users and permissions > Add user
E-mail: [your Google account]
Permission: Restricted

What I'll look at: which pages and queries already get impressions,
what changed over the last few months and any obvious indexing problems.
What I won't do: change any setting, submit anything or share your data.

If we don't work together, please remove me after [date, about 2 weeks
out], or I'll remind you to. Happy to do a 15-minute call instead if
you'd rather walk through it together.

[your name]

The last line is not filler. Some prospects will say no to access and yes to a screen share, and a screen share gets you most of the same five findings.

Pull these five findings in 30 minutes

This is a scan for three to five specific, checkable facts that make the proposal impossible to mistake for anyone else’s. Use the last 28 complete days, and compare against the 28 days before that, so both windows have the same length and the same weekday mix. Skip the most recent days, which can still be filling in.

1. Queries at positions 4-20 with impressions and few clicks. These are query and page pairs where Google already shows the site, usually below the top results. Sort by impressions, ignore branded navigational queries and keep the ones that match a page with commercial value. Our striking-distance guide covers the filtering in detail.

2. Pages that lost clicks between two equal windows. Look at the page dimension, compare the two windows and list the biggest losers. A page that dropped from, say, 900 to 400 clicks is a conversation starter. Before you put it in the proposal, rule out seasonality with a year-over-year check if the data goes back far enough. Our seasonality guide explains how.

3. Two pages competing for one query. Pick a query where two or more URLs each take a real share of impressions. Don’t call it a problem yet. Sometimes two pages serve different intents. In a proposal it works as a question: “Should /pricing or /plans own this query?” The cannibalization prompt has the checks.

4. Brand vs non-brand share. Split queries by the company name and its spellings. If, say, 85% of clicks come from people already typing the brand, the site has little organic reach outside its existing customers, and that is the honest headline. The method is in our brand vs non-brand guide.

5. Obvious indexing gaps. Open the Page indexing report, which restricted users can view. Look for important URLs (product, service, category pages) that are not indexed. Ignore the long tail of duplicates and redirects: Google’s own Page indexing help says duplicate or alternate pages shouldn’t be indexed. One key page missing from the index is worth more in a proposal than a count of 3,000 excluded URLs.

You won’t always find all five. Two strong findings with exact numbers are enough.

Doing it with an AI assistant

If you use an AI client connected through GSC Magic, the five pulls become one prompt. Once the prospect adds your Google account, their property appears in your account’s property list (gsc_list_sites returns every property the account can access, with your permission level), and you import it in the dashboard like any other site.

For the Search Console property [prospect’s property], use the last 28 complete days. Run report_quick_wins and give me the 10 non-branded query and page pairs with the most potential extra clicks. Run report_traffic_drops by page against the previous 28 days and list the 5 biggest losers with both click counts. Run report_cannibalization and list queries where two pages each get a meaningful share of impressions. Run report_brand_split with brand terms “[brand]”, “[brand misspelling]”. Show the exact numbers you used, mark anything that is a guess, and do not suggest fixes yet.

The tools return finished comparisons instead of raw rows, so you read a short list rather than a spreadsheet. report_index_health can add the indexing pass by reading the sitemaps and inspecting those URLs. If inspection calls fail on a restricted account, read the Page indexing report in Search Console instead.

Cost for a first pitch: every new account starts with a 7-day Pro trial with no card, which covers up to 10 sites. After that, the Free plan stays on: 1 site, all read tools and SEO reports, 200 requests a month and 10 a day, no write access. That is enough for one prospect at a time. Details are on the pricing page.

Turn findings into a range, not a promise

Never promise rankings. You don’t control Google, and a prospect who has been burned before knows that. What you can do is show the arithmetic behind an opportunity, with every assumption written down, and give a low and a high case.

The trick that makes the estimate defensible: take the click-through rate assumption from the prospect’s own data, not from an industry CTR curve. Their pages that already rank near the top show what their titles, their brand and their result features actually earn.

A worked example (illustrative numbers)

The numbers below are illustrative. They show the method, not a real client.

InputValueSource
Query and page pairs at positions 6-1514Performance report, last 28 days
Impressions on those pairs18,000Same
Clicks on those pairs270 (1.5% CTR)Same
CTR of their own pages ranking 2-46%Same, pages already near the top

Now two cases:

  • Low case. Only 30% of those impressions (5,400) come from pages that improve, and they reach half of the near-top CTR (3%). That is 162 clicks instead of 81, so about +80 clicks per 28 days.
  • High case. 60% of the impressions (10,800) come from pages that improve, and they reach the full 6%. That is 648 clicks instead of 162, so about +490 clicks per 28 days.

The proposal sentence is then: “Based on your current data, improving these 14 pages could add roughly 80 to 490 Google clicks per 28 days. We hold impressions flat, ignore seasonality and don’t attach a date, because ranking changes are not in anyone’s control.”

If they share a conversion rate from analytics, you can carry the range one step further. At an illustrative 2% conversion rate, that is about 1.6 to 9.7 extra leads per 28 days. Wide? Yes. That width is honest. A single number would be a guess dressed up as a forecast.

State the assumptions in the document, every one of them: flat impressions, the CTR source, which pages are included, and that average position is an average of the topmost result position, not a fixed rank everyone sees.

A proposal structure that fits on four pages

The structure follows the data. Each section answers one question the prospect will ask anyway.

  1. Your situation in your numbers. Three to five findings, each with a page or query, the numbers and the date range. No adjectives.
  2. Three priorities. Not twelve. One per finding that matters most, each with the evidence and what you would do.
  3. Scope. What you will deliver, in units the client can count: pages rewritten, templates fixed, internal links added, reports sent.
  4. Out of scope. Link building, paid search, development work, copywriting beyond the named pages. Whatever you don’t do, write it down.
  5. Timeline. What happens in weeks 1-2, 3-6 and after. Put the measurement points on it, not the outcomes.
  6. Price. A fixed fee for the defined scope, or a monthly retainer with the scope per month.
  7. How results will be reported. Which metrics, which comparison windows, how often, in what format. Our guide on showing SEO results to clients covers this part.

Here is the outline to copy:

SEO PROPOSAL: [Company], prepared [date]
Data: Google Search Console, [start date] to [end date],
compared with [previous start] to [previous end]

1. Where you are now
   - [Finding 1: page/query, numbers, window]
   - [Finding 2]
   - [Finding 3]
   - Brand share: [x]% of clicks come from branded queries

2. What we would fix first
   Priority 1: [page or issue] - evidence: [numbers] - work: [action]
   Priority 2: ...
   Priority 3: ...

3. Opportunity estimate (not a guarantee)
   Range: [low] to [high] clicks per 28 days
   Assumptions: [impressions flat, CTR source, pages included]

4. Scope
   Included: [countable deliverables]
   Not included: [explicit exclusions]

5. Timeline
   Weeks 1-2: [...]  Weeks 3-6: [...]  Review point: [date]

6. Price
   [fee, payment terms, notice period]

7. Reporting
   [metrics, comparison windows, frequency, format]

Access: we hold Restricted (view-only) access until [date]
and will remove it if we don't proceed.

When the prospect won’t share access

Some prospects will refuse, and that’s their right. It just means a different kind of proposal.

From public data you can still check real things: whether key pages return the right status, what the title and H1 say, whether robots.txt blocks something it shouldn’t, whether the sitemap lists the pages that matter, how the page renders. Third-party SEO tools give estimated rankings and traffic for public domains.

What you can claim changes. You can say “your /services page has no H1 and its title is cut off in search results”. You can’t say “your /services page lost 40% of its clicks”, because you don’t know that. Every number from a third-party tool goes into the proposal labelled as an estimate, with the tool named.

The practical fix is to make the first paid step a short data review: “Week 1: with access, we confirm or correct these findings against your Search Console and adjust priorities.” That turns the missing data into a scoped deliverable instead of a guess.

If they don’t sign, remove yourself

Access you no longer need is a liability for both sides. On the date you promised, remind them to remove you, or ask them to confirm they have. Google’s permissions page describes the steps: the owner opens Settings, Users and permissions, opens the menu next to your account and clicks Remove access.

Do your side too. Delete any exports you made, and if you imported the property into a tool, delete it there so no stored copy of their data stays in your account. In GSC Magic, that means removing the site from your dashboard.

It sounds small. It is also the kind of detail a prospect remembers when they come back six months later.

The proposal is the first deliverable

A proposal built on the prospect’s own Search Console data is already a piece of work, and the prospect can tell. It shows you can read their data, explain it plainly and estimate without overselling. That is the job they would be paying for.

If you sign the client, the next steps are the post-signing audit and a weekly routine for managing several clients. For freelancers running several pitches at once, the GSC Magic page for SEO freelancers covers the setup.

FAQ

Is it OK to ask a prospect for Search Console access before they sign?

Yes, if you ask for Restricted access, say exactly what you will look at, give a date when you will remove yourself and never touch settings. Restricted users can view performance data and most reports but cannot add users or submit sitemaps.

Should an SEO proposal include a traffic forecast?

Include an estimate, not a forecast. Show the inputs from their own data, state every assumption and give a low and a high case. Never promise rankings or a fixed number of clicks.

What if the prospect won’t share Search Console?

Write the proposal from public data, label every number as an estimate from a third-party tool or a public check, and make the first paid step a data review once access is granted.

How long should an SEO proposal be?

Short enough to read before the call. Two to four pages: their situation in numbers, three priorities, scope, exclusions, timeline, price and how results will be reported.

, Founder of GSC Magic

Łukasz builds GSC Magic, the MCP server that gives AI agents access to Google Search Console, Bing Webmaster Tools and GA4. He writes about using Search Console data with Claude, ChatGPT and coding agents, and about the SEO workflows those tools make possible.

lukaszblania.comgithub.com/lumizone

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.

Start free