The rank tool’s line jumped. Open Search Console first
On 24 September 2026 Search Engine Land wrote about google.com/goto on Google results. The piece describes it as a passthrough in front of the outbound link, tested from about July, aimed at scrapers such as SerpAPI. Earlier in the year, asking for a long list of results in one request was tightened, and a lot of rank-volatility sensors went noisy. A line that breaks, or a keyword that “falls out of the top 100,” is not by itself a demotion of your page. Often what broke is the tool’s own fetch.
A small firm rarely scrapes Google. The chart arrives in a monthly PDF: an agency or a rank tool says a core term moved from 6 to 30. If nobody checks that claim first, a redesign meeting gets booked. The person who owns the site should not change the title and should not rebuild the homepage yet. Open Search Console on the same query and the same dates as the PDF, and read impressions, clicks, and average position. If those three stay steady, the cliff is probably a miss in the scraper, not a buyer who cannot find you. If impressions, clicks, and average position move down together, then open that page and read the sentences.
The person comparing the two pictures uses the domain property. We use sc-domain so www and the bare host are not two piles of impressions. Use the whole string a buyer types, not a synonym the tool split off. Match the date range to the PDF. If the PDF is 28 days, Search Console is 28 days too. Do not switch it to 7 days and then compare the heights. Filter to Taiwan if that is who you serve. US impressions do not explain why the Chinese homepage gets no clicks. The reading method is in impressions without clicks. This note adds one rule: a rank scraped by a third party is not the first evidence for a redesign. The first evidence is your own Search Console.
Our September export is an example of how we read our own table, not a benchmark for another firm to copy. Few clicks, a few hundred impressions. Dozens of impressions on the romanized brand and almost no clicks. Impressions on “台北網站建置” with an average position outside the top twenty. There is no hidden “we are third” in that table. If a tool for the same dates says we are on page one for that query, the report disagrees with the export, and the export wins. Change how the report is read out loud. Do not rewrite a service page that already states the work, in order to “win back third.” That place was never in the export. We also do not promise that an edit will move the position forward.
Edit a page only for a reason that lives on that page. The first screen does not state the district and the work. The title does not match the body. Or you published a stack of unread drafts during a spam update. For the last one, wait out the rollout window in do not redesign for a two-week curve. A line that breaks inside a rank tool is not one of those three reasons. That broken line does not explain an empty form, and it does not tell you whether an overview invented your address.
With an agency, put one habit on the monthly call: the tool’s line cannot be the only reason to redesign. The PDF attaches a Search Console screenshot for that same query next to the tool’s line. If the two pictures do not agree on the height, the meeting talks about the Search Console numbers and whether the sentences on the page are clear. It does not talk about the color of the chart. Export the query table once a month and keep the file. That is cheaper than logging into a rank tool every day, and it keeps you from rewriting the site in the week the passthrough URL changed. We can read an export with you. We read the query table you exported, not a rank PDF that does not say which day it was scraped.
Two charts in the same meeting
The first chart in the meeting is a line that goes blank in mid-August. The note says “rank lost.” For that same query, Search Console still has impressions that week, and the average position is not the number written on the chart. Do not redesign this week because of that chart. Ask the person who writes the monthly report to say “the tool did not collect this query” next month, not “the site fell off page one.”
The other chart points the other way. Do not fold it into the blank line. It says a generic term is now in the top three, while Search Console shows zero impressions for that term. If impressions are zero and the tool still says top three, it did not measure your URL, or nobody searches that string for you. News that looks better gets the same check against Search Console before it enters a slide. A first place nobody saw is not this month’s revenue.
Questions
Is Search Console position the true rank?
It is the average position when your URL was shown for that query. It is not one laptop, in one city, at one minute. If you are deciding whether to edit the homepage this week, use that average, not the line a third-party scraper drew.
The tool says we fell off page one. Do we edit the homepage this week?
Do not edit yet. Open Search Console on the same query and the same dates as the tool. If impressions and average position did not fall together, leave the homepage alone this week. If they did fall, then look at whether fewer people searched that query, or whether the first screen of the service page never says what you do and where.
Do we buy a rank scraper?
Day to day we read Search Console exports, not a paid service that scrapes result pages. On the September table the romanized brand had impressions and few clicks, and the Taipei website-build query had impressions with a weak position. That is our own shape. It is not a reason for you to buy a scraper.
Read an export together Search and page structure
Source: Search Engine Land, 24 Sep 2026. Position comes from your Search Console, not from a third-party scrape.