Advanced: Find the Pages Worth Betting On With a No-Backlink Experiment

What You'll Get

An SEO experiment plan for a new site with zero backlinks and zero traffic: over four weeks, build a small set of pages around your product, let Search Console show you which demand is real, and then concentrate your effort on the pages that respond.

Throughout, this article follows a fictional "Markdown to PDF" micro-tool site, and includes two prompt templates, a symptom lookup table, and an experiment log template. The numbers are only a starting scale, not ranking requirements. AI helps organize data and revise drafts, but you make the call on which demand to pursue and whether the content is right.

Decide Three Things Before the Experiment

Focus on one product only. Every page should be able to naturally lead readers to your product; otherwise, when traffic arrives, you won't be able to serve it.
Keep the scope to 8 to 12 pages. Fewer makes patterns hard to see; more is more than one person can maintain. If real demand isn't there, make fewer pages rather than forcing it.
Judge results by data. After launch, look at Search Console, not your gut feeling.

Week 1: Find Keywords, Looking Only for "Small and Precise" Demand

If you have a keyword tool (Ahrefs, Semrush, or similar): set keyword difficulty (KD) to 0 to 5 and monthly search volume to 50 or below. Don't rush to delete terms that the tool shows as having zero search volume; numbers for small terms are often understated.

KD is the tool's estimate and is only useful for rough screening. Every candidate keyword should also pass two more checks:

Will people who search this use my product? Someone searching "markdown export pdf" will likely use it; someone searching "what is markdown" probably won't.
Can I write something better than what currently ranks at the top? It's only worth doing if you can fill a specific gap.

If you have no tool: look through questions customers have asked you, questions in user communities, and search autocomplete suggestions. If you don't know the search volume, leave it blank, and don't have AI estimate a number to fill in.

The following prompt turns candidate keywords into an actionable page list:

New accounts get 10 free credits, plus 1 more for each daily check-in. Spend 1 credit to unlock and it stays open for you.

Week 2: Check Search Results and Decide What Page Each Term Becomes

Search each group's representative term in Google for your target market, look at what the top 5 to 10 organic results mostly are, and then act accordingly:

Mostly online tools: build a tool page that works as soon as it opens, with explanatory text below the tool.
Mostly tutorials: write steps, and add "how to confirm it worked" and "how to troubleshoot if it failed."
Mostly comparisons and reviews: explain "which one to choose in which situation," and state the limitations clearly.

Using the sample site as an example: if the results for "markdown to pdf online" and "export md file to pdf" look similar, serve both with the same tool page; "markdown to pdf fonts broken" is a different problem and suits a separate troubleshooting tutorial. (The sample terms only illustrate the approach; their real search volumes have not been checked.)

Week 3: Write the Pages and Weave Them into a Network

The structure is one core page plus several supporting pages. In the sample site, the core page is the conversion tool itself, and the supporting pages each solve specific problems such as garbled Chinese characters, pagination, code highlighting, and misaligned tables.

Every page follows three rules:

Give the answer on the first screen. State the conclusion or offer the tool directly, and put background afterward.
Where there are steps, include troubleshooting. Describe the two or three most common failures and how to fix them.
Put links at the "next step." After a troubleshooting tutorial explains the cause, link back to the tool page on the line "convert again." Links exist to serve readers, and there is no target for their number.

From Week 4 On: Review the Data Weekly Against the Symptom Table

Open Search Console. The first thing to do is use URL Inspection to see whether the page has been indexed; if it has, go to the Performance report, filter by page, and see which queries it appears for.

Symptom	Most likely cause	What to do first
URL Inspection shows not indexed	Crawling or quality issue	Read the specific reason given in the inspection result and address each one
Indexed, almost no impressions	The demand chosen is too cold, or the page format is wrong	Go back to Week 2 and compare against the search results again
Has impressions, average position 11 to 30	The content doesn't fully answer the question	Find related queries where the body doesn't answer well yet, and add them to the body
Ranking isn't bad, but few clicks	The title and description don't make the benefit clear	Revise the title, and search for it yourself to see how the result actually displays

Example: if "converted image not showing" starts getting impressions, add the cause and before-and-after screenshots to the body, rather than just stuffing those words into the title.

Habits for reading the numbers: if impressions are too few, extend the date range to 28 days before looking; when comparing before and after, keep country, device, and time span consistent; average position changes with the query and does not represent a fixed rank.

For how to fill in gaps section by section on pages that already rank, see Health Check for Pages That Already Rank.

Revision Prompt: Have AI Change Only What Needs Changing

Attachments: The body text of [URL], and its Search Console query export.

Please do the following in order:
1. Find queries that have impressions, are related to the page's topic, but are not well answered in the body; prioritize those with average position 11-30;
2. For each query, explain which paragraph it should be added to, and provide the rewritten paragraph; keep unmentioned paragraphs unchanged;
3. List what I need to supply myself: screenshots, test results, citation sources;
4. If the data is insufficient to judge, tell me directly what is still missing, and do not guess.

Experiment Log Template

Log one line for each change, so that during review you can tell which cut made the difference:

URL	Problem before the change	What changed	Date	Impressions after 2 weeks	Clicks after 2 weeks	Conversions
/markdown-to-pdf	Ranking 18, garbled-character query not answered	Added a "Chinese garbling" section and a comparison image	11-08

For important pages with substantive changes, you can request re-indexing in URL Inspection; repeated submissions won't make it faster, and they don't guarantee indexing.

After the First Round

Roughly once a quarter, do a review: look at impressions, clicks, and real usage in the product together, and decide what the next group of pages should cover. This sets the rhythm of review; it is not a promise of how long results will take.

Backlinks can accumulate slowly alongside this. Focus on relevant resource pages and industry sites that would genuinely cite you; fewer and relevant beats many and scattered.

Notes

Search volume and difficulty with no tool data behind them are treated as unknown.
Better fewer pages than thin ones: 10 carefully written pages are more useful than 50 empty shells.
A new site takes time from indexing to showing impressions, so very little data in the first few weeks is normal.