How often does Sempuls check my keyword rankings?

How often we check depends on your plan. Free trial and Starter: once every 7 days, Standard: every 3 days, Pro and Agency: daily. You can also run a manual check at any time.

Does Sempuls detect AI Overview in Google?

Yes — Sempuls automatically detects the presence of AI Overview (AIO) for each keyword. You can check whether your domain is cited in the AI answer.

Will social media activity alone improve rankings?

Indirectly it can help – if it generates traffic, links, and brand trust.

How long does it take to see results from increased SEO activity?

The first results may be visible after a few weeks, but the biggest advantage comes from consistency over months and years.

Can a lack of activity lower your ranking?

Yes – a business that stops being active gradually loses quality signals and visibility.

Are free tools enough to track rankings?

Yes, Google Search Console and manual checks can give you a picture of your rankings, but they aren't as convenient or complete as paid tools.

How often should I check my site's ranking?

Ideally daily or several times a week – changes in Google's algorithms can happen quickly.

Is manual searching reliable?

Only if you use incognito mode and turn off cookies and personalization parameters – otherwise the result may be skewed.

Which ranking monitoring tool do you recommend?

It depends on your needs and budget – for most businesses SEMrush or Surfer will be a good choice, and to start it's worth using the free Ahrefs Webmaster Tools and GSC.

What is Local Grid and how does it work?

Local Grid is a grid of geolocation points around your business. Sempuls checks your Google Maps Business Profile ranking at each grid point, showing your visibility range on a heatmap.

Can I monitor multiple locations at once?

Yes — on the Pro plan you can monitor up to 30 Google Business Profiles, and on Agency up to 50. Each location can have its own Local Grid.

Is recycling an article into a GBP post considered duplicate content by Google?

No, as long as the content is actually rewritten, not copied. A Google Business Profile post and a blog article differ in length, tone, structure and audience – Google treats them as two separate documents. A problem only arises if you literally copy fragments of 1:1 – then Google Business Profile may reject the post as spam, and Google may lower the page's rank. Good AI that generates a post from an article URL always rewrites rather than summarizes.

How often can I publish GBP posts from the same article?

At most once every 6 months, and only if the article has been updated or the context has changed (e.g. seasonality). Repeating a post from the same source at shorter intervals is a thin content signal for AI and discourages your profile's followers. That's why source tracking ("Used") is key – without it you'll lose track after just 30-50 posts.

How many GBP posts should come from one blog article?

One article = one Google Business Profile post. Trying to squeeze 2-3 posts out of the same article leads to redundancy and cannibalization of AEO signals. It's better to write two shorter articles than one long one if you plan to recycle.

Does AI that generates GBP posts from articles require a separate model subscription?

It depends on the tool. Sempuls uses built-in access to Claude Haiku – generating one post costs about PLN 0,008, so even with 200 posts a month the expense is negligible. Some cheaper tools use free alternatives, but the quality of prompts and rewriting is then lower.

How do I measure the effectiveness of blog → GBP recycling?

Three metrics give the full picture: the number of Google Business Profile views in "Insights", the number of clicks on "Website" from the profile (Google Search Console will show traffic from domain.com/blog/* by referrer), and your business's position in the Local Pack for your main local keywords. First results show after 4-6 weeks, full results after 3-6 months of consistent publishing.

Does this mechanism also work for businesses without a physical location?

Only partially. GBP profiles can only be created for businesses that serve customers face to face or in a specific geographic area (service-area business). If you run only an e-commerce store shipping packages across Poland, blog → GBP recycling doesn't apply. In that case, it's better to focus on blog → social media and blog → newsletter recycling.

What if the client doesn't have a company blog yet?

Start with the blog, not Google Business Profile. Without systematic blog content production there's nothing to recycle, and writing 2-3 Google Business Profile posts a week from scratch doesn't scale. A cheaper investment is building a backlog of 20-30 good blog articles – that gives you material for 6 months of Google Business Profile posts.

What is AEO (Answer Engine Optimization)?

AEO is optimizing content for AI answer engines — such as Google AI Overviews, ChatGPT, Perplexity, and Gemini. Sempuls analyzes whether your content meets the criteria for being cited by AI.

What are AI buttons and can I add them to my business website?

AI buttons are buttons on a business website that, when clicked, open a language model (ChatGPT, Perplexity, Claude, Gemini, and others) with a ready-made question about your business. Each button points to a different AI model and leads to a chat interface with a pre-filled prompt containing the business name, category, address, and a backlink. You can add them to any business website — in Sempuls they are generated automatically on every business profile page.

Do AI buttons really affect how ChatGPT and Perplexity see my business?

Yes, they work on three levels. First, each click generates a session in an AI model with a specific question about your business — the model opens your site and reads its content. Second, users often share these answers further (screenshots, copying to social media), which builds brand mentions across the internet. Third, in the long term the brand appears in the training corpus of subsequent model versions — and then AI starts citing it even without fetching the page.

Why do I need as many as 10 different AI buttons? Isn't ChatGPT enough?

Every language model has its own training corpus, its own search mechanism and its own user base. ChatGPT dominates the general segment, but Perplexity is preferred by professionals looking for sources, Claude is chosen by people who value accuracy, Gemini dominates on Android, Copilot is preinstalled in Windows, Grok is growing on X, and Meta AI integrates with WhatsApp and Instagram. The fact that ChatGPT knows your business doesn't mean Claude or Mistral does. 10 buttons means 10 independent chances of being cited.

Are AI buttons safe, and won't they hurt my classic SEO?

They are safe. All links have rel="nofollow noopener noreferrer" attributes — they don't pass SEO value to AI platforms and open in a new tab, so the user doesn't lose their session on your site. Adding AI buttons doesn't change the page's structure or content, so classic SEO stays untouched. On the contrary — signals generated by higher engagement and new external links (from shared AI answers) also strengthen rankings in classic Google results.

What exactly should the prompt sent to the AI contain?

An optimal prompt contains three elements. First: a clear instruction to describe a specific business ("Introduce company X" works better than "Tell me something about this company"). Second: key identifying data — the name, business category, address. Third: a request to cite and name the source ("Use quotes from the content and give the source: https://twojafirma.pl"). The optimal prompt length is 2-3 sentences, up to about 250 characters after encoding.

How can I measure the results of implementing AI buttons on my site?

Three metrics are worth tracking in the first 3 months. First: the number of button clicks (measured in Google Analytics 4 or your own tracker) — it shows which model your users choose. Second: referral traffic from AI domains (chat.openai.com, perplexity.ai, claude.ai) as a referrer in GA4 — a sign that models have started citing your site. Third: direct brand mentions in AI answers — manual checking or tools like Sempuls AI Citations, which monitor brand appearances across 8 models at once.

Will AI buttons work if my business only has a Google Business Profile and no website of its own?

Unfortunately, no. AI buttons require a page to place them on — Google Maps alone isn't enough, because language models don't have access to the Google Maps database and can't "fetch" your profile. The good news: in Sempuls every business in the database automatically gets a generated business page (e.g., sempuls.com/your-company) with built-in AI buttons. It's an ideal solution for small businesses that don't have their own website but want to be visible to AI.

Will AEO replace SEO?

No. AEO is a layer on top of SEO. Without good SEO your site won't make it into retrieval, so AI will have nothing to cite. AEO adds optimizations at the structure and schema level that increase the chance of being selected from the index.

How long does AEO implementation take?

Technical changes (schema, H2/H3 structure, BLUF, llms.txt) – 2–4 weeks for an average site. First citations in AI Overviews – 4–8 weeks after indexing. Steady growth – 3–6 months.

Are AI Overviews stealing my traffic?

Sometimes yes – for informational queries CTR can drop by 30–60%. But for transactional and local queries traffic grows, because AI directs people to specific offers. Strategy: focus on keywords with purchase intent.

Do I need to block GPTBot if I don't want to be cited?

You can – via robots.txt. But then you lose free brand exposure. Most companies in 2026 deliberately allow AI bots, because a citation with a link is free PR.

Can I optimize for a specific engine (e.g. only ChatGPT)?

Partly. Each engine has its own preferences (Perplexity loves tables, ChatGPT – lists, AI Overviews – short definitions). But 80% of optimization is shared and works everywhere.

What is the most common mistake when implementing AEO?

Writing overly long, "everything about…" articles. AI doesn't cite 3000-word threads. It cites specific, short answers. Instead of one mega-article, it's better to have 5 focused answers on separate URLs.

Will AI Overviews replace traditional SEO?

No, but it changes how you optimize content – more detail is needed.

Can any domain be cited?

Yes, if it provides the best-quality answers on a given topic.

Does AI Overviews always show a link?

Most of the time, yes – which is why it's worth optimizing your copy for that goal.

How quickly can you get cited?

Sometimes after just a few content updates and implementing structured data.

What is the Generative AI report in Google Search Console?

It's a dedicated view in Search Console, launched on June 3, 2026, that shows how often your pages appear in Google's generative features – AI Overviews and AI Mode in search and in Discover. You'll find it under Performance → Search results.

Does the report show clicks and CTR?

No. The first version shows only impressions, pages, countries, devices and dates. Data on clicks, click-through rate and queries is missing. Google has announced that more metrics will be added in the future.

Why don't I see the Generative AI tab in my dashboard?

The feature is rolling out in stages – first to a subset of site owners in the UK, then globally. Polish sites may not have access yet. Check Search Console periodically, because the reach will keep expanding.

Can I opt my content out of AI Overviews?

Yes. Google is testing a switch that lets you opt out of AI features. Opting out means no impressions or traffic from those features, but according to Google it does not lower your position in regular search results. For most businesses, we recommend staying in.

Can I download this data via the API?

Not yet. The report lives only in the Search Console interface – the API (searchanalytics.query) doesn't yet have a separate type for AI Overviews or AI Mode. You can only export the data to CSV manually. Historically, Google's API catches up on such reports a few months after they launch in the dashboard.

Does SEO still make sense if Google answers through AI?

Yes. Google's generative features rely on the same index and ranking systems as classic results. Clear content, structured data and strong E-E-A-T signals help you get cited by AI in exactly the same way they help in traditional search.

Does phone click tracking work on all devices?

It works on most devices, although some browsers may block some events.

Do I need to use GTM to track phone clicks?

No, but GTM makes implementing and managing events much easier.

Does a click on a number mean the user actually made the call?

Not always – a click initiates the call, but you can't be sure the conversation actually took place.

Does click tracking affect SEO?

Indirectly, yes – the data helps you improve UX and conversions, which affects the quality of your site.

False ranking drops in Google - poisoned SERPs 2026

Since 24 September 2026, Google has been serving some of the bots that check rankings result pages that look normal but have their organic listings swapped out. Anyone who reads a rank report sees the effect: a keyword suddenly falls out of the top 30, even though the site has lost nothing in Google. Here we show what public sources say about this, what Sempuls data showed on 2 November 2026, and how to tell a real drop from a measurement error.

What has been happening to Google results since late September

Rank tracking tools don't sit in front of a browser. They fetch Google result pages automatically, thousands of times a day, and read your domain's position from them. Since 24 September, Google has been recognising these fetches more and more often. Instead of the real result page, it then returns one of two versions:

a page with swapped results - with ads and an AI summary, but with a different set of pages in the organic results, often without company websites and with aggregators and directories instead,
a cut-down page - with only a few results or no further pages, so a keyword ranking 15th has nowhere to appear.

The industry calls these poisoned SERPs. Google has made no statement about it. According to AccuRanker, the decision is made for a whole browser session, not for a single keyword, so one poisoned session can spoil many measurements at once.

What public sources say

Several companies that measure rankings themselves have already described the problem. The key data:

AccuRanker (article from 19 October, updated until 26 October) reports that before the change whole result pages flipped in about 1 in 1,000 checks. On 25 September the share of poisoned desktop pages reached 30%, then fell to 1-3% a day. On 24 October Google poisoned 40% of first fetches, and after checks and refetching the bad ones, 3.2% remained. Google tightened the mechanism again on 15, 20 and 23 October.
Nozzle, using its own method, showed a peak of 27% poisoned results on 29 September.
stat4seo described Polish result pages without company websites as early as 6 October.
DataForSEO, in a notice from 23 October, did not announce a fix, only an extra field for verifying the result.
cloro, in its comparison up to 31 October, showed that some popular Google results APIs returned a full page in about 10% of attempts, and the best ones in over 95%. The same source reports another change on Google's side on the evening of 26 October.

One conclusion follows from these sources: the problem is real, it affects the whole industry and no tool is immune to it. Tools differ in what they do with a bad result.

What we see in Sempuls data

Sempuls measures around 900 client keywords on the Polish market every day. This shows how big the problem is in Polish results, which foreign reports barely cover.

Fewer hits in the first measurement

Before 24 September, the nightly measurement found the client's domain in the results for about 890 keywords. By mid-October it was only about 550, even though client sites were not losing traffic from Google. At the turn of October and November, 52-61% of keywords that had been in the top 10 the day before disappeared from the results in the first measurement. That is far more than the few percent reported by trackers that reject poisoned pages themselves.

A second check is no longer enough

We check every lost keyword again with a different tool. Between 20 and 26 October, this recheck recovered about 60% of keywords. After the change on 26 October, its success rate dropped to 20-25%. This matches the moment other companies reported another tightening on Google's side.

A test of 266 lost keywords

On 2 November we took all keywords from that day that two measurement tools had failed to find and checked them with a third provider. The result:

in 78% of cases the client's site was in Google's results as usual,
excluding keywords that had already been far outside the top 30, in 86%,
positions differed from the last reliable measurement by 1-2 places on average.

So most of that day's drops were measurement errors, not problems with the site.

How much Google Search Console data is worth

Search Console is not poisoned, because it shows positions from real impressions. It has two limits, though. Data arrives with a 2-3 day delay, and for keywords the site got no impressions for, there is no data at all. We checked how accurate different ways of filling the gaps are, comparing each with the next real measurement:

the last reliable position confirmed by impressions in Search Console - 72% within two places,
carrying over the last reliable position without confirmation - 71%,
the average position from Search Console alone - only 35%.

Search Console is good at confirming that a site is still in the results. As a standalone source of the exact position it performs poorly, because an average of a few impressions can be off by several places.

How to spot a false ranking drop

Before you start fixing your site after a drop in a report, check a few things:

Does the drop affect a single day? According to AccuRanker, more than 9 in 10 poisoned results for a keyword last one day. If the keyword is back the next day, it was a measurement error.
Did many keywords disappear at once? A sudden drop of dozens of keywords on one day, with no change to the site, is the typical picture of poisoned results.
Did your competitors disappear with you? If the saved results are also missing pages that usually sit next to yours, suspect the data, not the site.
What does Search Console say? Filter by keyword and date. If the average position and the number of clicks did not change, the drop in the tool was false.
Are there businesses from another town in the local results? This is a common sign of a swapped result page.

Don't react to a single day. Base decisions about changes to your site on the trend across several measurements.

What we do at Sempuls

Since early October, rank monitoring in Sempuls has had a separate path for days when measurement is disrupted:

Rechecking lost keywords with a different tool, in a different session and at a different time.
Comparison with Google Search Console. If the site has impressions and the last reliable measurement was recent, we show that position as filled in, not as a drop.
Clear labels in the rank calendar. Disrupted days and filled-in positions are marked, so you know what is a measurement and what is an estimate.
Testing an additional results provider. Following the results from 2 November, we are testing a source that found most of the lost keywords in our test.

We are also working on a check before a measurement is saved. If the new top 10 for a keyword shares fewer than half of its pages with recent days, the result will be fetched again before it reaches the report. AccuRanker describes this approach. According to their data, a single refetch returns a normal result in roughly 3 out of 4 cases, and further attempts and checks bring the share of bad results down to a few percent.

Summary

Since 24 September 2026, Google has been deliberately making automated rank checking harder, and since 26 October it has been doing so even more effectively. In Sempuls data from the Polish market, most sudden drops are measurement errors. In the test from 2 November, 78-86% of lost keywords were in the results as usual. Don't draw conclusions from a single day, check Search Console, and use a tool that verifies suspicious drops itself before showing them in a report.

This website uses cookies